Electronic device, recording medium, and method for clustering data related to battery charging pattern thereof

Abstract

An electronic device for clustering data related to a battery charging pattern is provided. The electronic device can acquire a plurality of pieces of battery charging data for a plurality of vehicles, generate a plurality of feature maps related to a battery charging pattern on the basis of each of the plurality of pieces of battery charging data, perform image processing on each of the plurality of feature maps on the basis of an image pooling technique, calculate similarity between each of the plurality of feature maps on which the image processing has been performed, determine, on the basis of the calculated similarity, a clustering algorithm and an index for evaluating a clustering result, the optimal number of clusters for the plurality of feature maps on which the image processing has been performed, and cluster the plurality of feature maps on which the image processing has been performed into the optimal number of clusters.

Description

Method for clustering data relating to electronic devices, recording media and their battery charging patterns

[0001] The present disclosure relates to a method for clustering data regarding an electronic device, a recording medium, and a battery charging pattern thereof, and more particularly, to a technique for clustering battery charging pattern data by setting boundaries between types regarding the charging patterns of electric vehicle users to be more clearly distinguished.

[0002] This application claims the benefit of priority to Republic of Korea Patent Application No. 2024-0109373, dated August 14, 2024, and Republic of Korea Patent Application No. 2024-0144730, dated October 22, 2024, the entire contents of which are incorporated herein by reference.

[0003] To identify charging patterns of electric vehicle users, time-series data on battery charging patterns can be processed and clustered to standardize charging and discharging characteristics without distortion. However, excessive clustering of data can lead to increased data diversity and the number of outliers that are difficult to standardize. This can blur the boundaries between different types. Unsupervised learning clustering techniques can be utilized to address this issue, but data can still exist at the boundaries between clusters, making it difficult to find distinct boundaries between clusters.

[0004] To address this, density-based clustering techniques like DBSCAN (Density-Based Spatial Clustering of Applications with Noise) exist. However, these techniques require a large number of hyperparameters to be adjusted based on the characteristics of the target data, and clustering performance can be highly dependent on these hyperparameters. Furthermore, when there is an excessive amount of data, there may be a problem in that distinguishable density differences cannot be guaranteed.

[0005] Therefore, conventional approaches have involved designating a sufficiently large number of clusters and manually reducing them one by one. However, this approach inevitably involves subjective elements, is time-consuming, and consumes significant resources. Furthermore, because there is no clear evidence that the final, artificially reduced clusters are optimal, the evaluation of clustering performance is ambiguous. Consequently, the need for a method that can identify an optimal number of distinct types from a significant amount of data has emerged.

[0006] An embodiment of the present disclosure is proposed to solve the above-described problem, and provides a method for clustering data related to an electronic device and its battery charging pattern.

[0007] The technical task to be achieved by this embodiment is not limited to the task described above, and other technical tasks can be inferred from the following examples.

[0008] An electronic device according to one embodiment includes a transceiver; a processor; and one or more memories storing one or more instructions, wherein the one or more instructions, when executed, cause the processor to obtain a plurality of battery charging data for a plurality of vehicles, generate a plurality of feature maps regarding a battery charging pattern based on each of the plurality of battery charging data, perform image processing on each of the plurality of feature maps based on an image pooling technique, calculate a similarity between each of the plurality of feature maps on which image processing has been performed, determine an optimal number of clusters for the plurality of feature maps on which image processing has been performed based on the calculated similarity, a clustering algorithm, and an index for evaluating a clustering result, and perform clustering of the plurality of feature maps on which image processing has been performed into clusters having the optimal number of clusters.

[0009] According to one embodiment, the plurality of battery charging data may include information on the state of charge (SoC) of the batteries over time, as data collected for each of the plurality of batteries included in the plurality of vehicles over a preset period of time.

[0010] According to one embodiment, one or more instructions may be configured to cause the processor, when executed, to: identify one or more charging sections based on first battery charging data for a first vehicle among a plurality of battery charging data for a plurality of vehicles; identify, for each of the one or more charging sections, an SoC of the battery at a charging start time and an SoC of the battery at a charging end time; and generate a first feature map indicating a frequency of the one or more charging sections corresponding to a range of the SoC of the battery at the charging start time and a range of the SoC of the battery at the charging end time.

[0011] In one embodiment, the frequency of one or more charging intervals can be determined based on a probability density function.

[0012] According to one embodiment, the image pooling technique may include a max pooling technique that selects a maximum value corresponding to a plurality of regions included in a target feature map and generates a feature map composed of the maximum values ​​of the plurality of regions, and a dilation pooling technique that generates a feature map composed of representative values ​​of at least some values ​​included in each of the plurality of regions.

[0013] According to one embodiment, one or more instructions, when executed, may be configured to cause the processor to compute a similarity between each of a plurality of feature maps on which image processing has been performed, based on a structural similarity index.

[0014] According to one embodiment, one or more instructions may be configured to cause the processor, when executed, to perform clustering multiple times for a plurality of feature maps on which image processing has been performed, with different numbers of clusters according to the calculated similarity and the clustering algorithm, and to determine an optimal number of clusters for the plurality of feature maps on which image processing has been performed based on the clustering results performed multiple times and an index for evaluating the clustering results.

[0015] In one embodiment, the clustering algorithm may include a hierarchical clustering algorithm that hierarchizes at least some of the feature maps included in the plurality of feature maps on which image processing has been performed based on similarity.

[0016] According to one embodiment, the index includes a first index, a second index, and a third index, and one or more instructions may be configured to cause the processor, when executed, to calculate, based on clustering results performed multiple times, a first index according to the number of clusters, a second index according to the number of clusters, and a third index according to the number of clusters, respectively, and determine an optimal number of clusters for a plurality of feature maps on which image processing has been performed based on a first number of clusters corresponding to a maximum or minimum value among the first indices according to the number of clusters calculated, a second number of clusters corresponding to a maximum or minimum value among the second indices according to the number of clusters calculated, and a third number of clusters corresponding to a maximum or minimum value among the third indices according to the number of clusters calculated.

[0017] According to one embodiment, each of the first index, the second index, and the third index may include at least one of a Silhouette Index technique, a Calinski-Harabasz Index technique, and a Davies-Bouldin Index technique that evaluate clustering results based on the cohesion of each cluster and the separation between clusters, for each clustering result performed according to a clustering algorithm.

[0018] A method for clustering data regarding a battery charging pattern according to one embodiment may include the steps of: obtaining a plurality of battery charging data for a plurality of vehicles; generating a plurality of feature maps regarding the battery charging pattern based on each of the plurality of battery charging data; performing image processing on each of the plurality of feature maps based on an image pooling technique; calculating a similarity between each of the plurality of feature maps on which image processing has been performed; determining an optimal number of clusters for the plurality of feature maps on which image processing has been performed based on the calculated similarity, a clustering algorithm, and an index for evaluating the clustering result; and performing clustering on the plurality of feature maps on which image processing has been performed into clusters having an optimal number of clusters.

[0019] A computer-readable, non-transitory recording medium having recorded thereon a program for executing a method for clustering data regarding a battery charging pattern according to one embodiment of the present invention on a computer, the method for clustering data regarding a battery charging pattern comprising: obtaining a plurality of battery charging data for a plurality of vehicles; generating a plurality of feature maps regarding the battery charging pattern based on each of the plurality of battery charging data; performing image processing on each of the plurality of feature maps based on an image pooling technique; calculating a similarity between each of the plurality of feature maps on which image processing has been performed; determining an optimal number of clusters for the plurality of feature maps on which image processing has been performed based on the calculated similarity, a clustering algorithm, and an index for evaluating a clustering result; and performing clustering of the plurality of feature maps on which image processing has been performed into clusters having an optimal number of clusters.

[0020] According to the present disclosure, it is possible to discover major types of battery charging patterns even from multiple battery charging data with unclear distinguishability.

[0021] In addition, according to the present disclosure, by emphasizing the characteristics of each data and improving the distinguishability between data, the boundaries between each cluster can be highlighted, ultimately making clustering easier to perform.

[0022] In addition, according to the present disclosure, by selecting an appropriate number of clusters, it is possible to discover a type of charging pattern that can more efficiently identify the characteristics of data.

[0023] The effects of the invention are not limited to the effects mentioned above, and other effects not mentioned will be clearly understood by a person skilled in the art from the description of the claims.

[0024] Figure 1 illustrates a block diagram of an electronic device according to one embodiment.

[0025] FIG. 2 illustrates a flowchart of a method for clustering data regarding battery charging patterns according to one embodiment.

[0026] FIG. 3 illustrates a process for generating a feature map based on data regarding a battery charging pattern according to one embodiment.

[0027] Figure 4 shows data regarding a battery charging pattern according to one embodiment.

[0028] Figure 5 shows statistical data regarding a charging section according to one embodiment.

[0029] Figure 6 shows a probability distribution graph regarding the frequency of charging intervals according to one embodiment.

[0030] Figure 7 illustrates a feature map according to one embodiment.

[0031] Figure 8 illustrates an image processing process based on a maximum pooling technique according to one embodiment.

[0032] Figure 9 illustrates an image processing process based on an extended pooling technique according to one embodiment.

[0033] Figures 10a to 10c illustrate multiple feature maps on which a pooling technique is performed according to one embodiment.

[0034] Figure 11 illustrates a clustering process based on similarity between images according to one embodiment.

[0035] Figure 12 illustrates a clustering process based on a clustering algorithm and index according to one embodiment.

[0036] Figure 13 illustrates a process for determining the optimal number of clusters according to one embodiment.

[0037] Figure 14a shows a first index according to the number of clusters according to one embodiment.

[0038] Figure 14b shows a second index according to the number of clusters, according to one embodiment.

[0039] Figure 14c shows a third index according to the number of clusters, according to one embodiment.

[0040] The terms used in the examples have been selected from widely used, current terms, taking into account the functions of the present disclosure. However, these terms may vary depending on the intentions of those skilled in the art, precedents, the emergence of new technologies, etc. Furthermore, in certain cases, terms may be arbitrarily selected by the applicant, in which case their meanings will be described in detail in the relevant description. Therefore, the terms used in this disclosure should not be defined simply as names, but rather based on the meanings of the terms and the overall content of the present disclosure.

[0041] When a part of the specification is said to "include" a component, this does not exclude other components, but rather implies the inclusion of other components, unless otherwise specifically stated. Furthermore, terms such as "part" and "module" used in the specification mean a unit that processes at least one function or operation, which may be implemented in hardware, software, or a combination of hardware and software.

[0042] The expression "at least one of a, b, and c" described throughout the specification may encompass 'a alone', 'b alone', 'c alone', 'a and b', 'a and c', 'b and c', or 'all of a, b, and c'.

[0043] Below, embodiments of the present disclosure are described in detail with reference to the attached drawings so that those skilled in the art can easily implement the present disclosure. However, the present disclosure may be implemented in various different forms and is not limited to the embodiments described herein.

[0044]

[0045] Hereinafter, embodiments of the present disclosure relating to an electronic device for clustering data regarding a battery charging pattern are described in detail with reference to the drawings.

[0046] Figure 1 illustrates a block diagram of an electronic device according to one embodiment.

[0047] Referring to FIG. 1, an electronic device (100) may include, according to one embodiment, a transceiver (110), a processor (120), and a memory (130). The electronic device (100) illustrated in FIG. 1 only includes components related to the present embodiment. Therefore, it will be understood by those skilled in the art related to the present embodiment that other general components may be included in addition to the components illustrated in FIG. 1.

[0048] For example, the electronic device (100) may include a communication device including one or more transceivers (110), an input unit, and an output unit. The communication unit is a device for performing wired / wireless communication and may communicate with an external electronic device. The external electronic device may be a terminal or a server. In addition, communication technologies used by the communication unit may include GSM (Global System for Mobile communication), CDMA (Code Division Multi Access), LTE (Long Term Evolution), 5G, WLAN (Wireless LAN), Wi-Fi (Wireless-Fidelity), Bluetooth, RFID (Radio Frequency Identification), Infrared Data Association (IrDA), ZigBee, NFC (Near Field Communication), etc. The input unit may be, for example, a traditional keypad or keyboard, a mouse, a microphone for receiving voice signals, a camera, and various other input means for detecting or receiving various types of user input. The output unit may be, for example, a display that outputs images, a speaker that outputs sounds, a haptic device that generates vibrations, and various other forms of output means.

[0049] According to one embodiment, the electronic device (100) may be a server that acquires and processes data for multiple vehicles. Specifically, the data for multiple vehicles may include multiple battery charge data for the multiple vehicles. The multiple battery charge data may be acquired, for example, from at least one of an On-Board Diagnostics (OBD) device mounted on each vehicle, a battery management system (BMS), and a device (e.g., a database) in which battery charge data is previously stored, via the transceiver (110) of the electronic device (100). The manner in which the electronic device (100) acquires the multiple battery charge data is not limited to the above example, and it will be clearly understood by those skilled in the art that the electronic device (100) may acquire the multiple battery charge data from various devices with which it can communicate via the transceiver (110). The type of the electronic device (100) is not limited thereto, and various embodiments of the present disclosure may be applied to various devices capable of acquiring and processing data for vehicles.

[0050] The processor (120) can control the overall operation of the electronic device (100) and process data and signals. The processor (120) can be composed of at least one hardware unit. In addition, the processor (120) can operate by one or more software modules generated by executing program codes stored in the memory (130). The processor (120) can include a memory, and the processor (120) can control the overall operation of the electronic device (100) and process data and signals by executing the program codes stored in the memory.

[0051] The processor (120) may be implemented as a computer or similar device based on hardware, software, or a combination thereof. In terms of hardware, the processor (120) may be implemented in the form of an electronic circuit that processes electrical signals to perform control functions, and in terms of software, the processor (120) may be implemented in the form of a program that drives the hardware processor (120). Meanwhile, unless otherwise specified in the following description, the operation of the electronic device may be interpreted as being performed under the control of the processor (120). That is, when modules implemented in the clustering system for data regarding battery charging patterns are executed, the modules may be interpreted as controlling the processor (120) to perform the following operations of the electronic device (100).

[0052] The memory (130) can store various types of information. The memory (130) can store data temporarily or semi-permanently. For example, the memory (130) of the electronic device (100) can store data related to an operating program (OS: Operating System) for operating the electronic device (100). Examples of the memory (130) may include a hard disk drive (HDD: Hard Disk Drive), a solid state drive (SSD), flash memory, read-only memory (ROM: Read-Only Memory), random access memory (RAM: Random Access Memory), etc. The memory (130) may be provided as a built-in type or a detachable type.

[0053] In summary, the various embodiments may be implemented through various means. For example, the various embodiments may be implemented through hardware, firmware, software, or a combination thereof.

[0054] In the case of hardware implementation, the methods according to various embodiments may be implemented by one or more application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), processors, controllers, microcontrollers, microprocessors, etc.

[0055] When implemented via firmware or software, the methods according to various embodiments may be implemented in the form of modules, procedures, or functions that perform the functions or operations described above. For example, software code may be stored in memory and executed by a processor. The memory may be located within or external to the processor and may exchange data with the processor via various known means.

[0056] FIG. 2 illustrates a flowchart of a method for clustering data regarding battery charging patterns according to one embodiment.

[0057] In step S210, the electronic device (100) may acquire multiple battery charging data for multiple vehicles. Here, the multiple battery charging data is data collected for each of the multiple batteries included in the multiple vehicles over a preset period of time, and may include information on the state of charge (SoC) of the batteries over time. The preset period may be a period during which battery charging and discharging may be repeated multiple times, and may be, for example, one month. Accordingly, the multiple battery charging data may be time-series data including information on changes in SoC over time during multiple repetitions of charging and discharging of each of the multiple batteries included in the multiple vehicles. A more specific embodiment of the battery charging data will be described in detail below with reference to FIG. 4.

[0058] In step S220, the electronic device (100) may generate a plurality of feature maps regarding the battery charging pattern based on each of the plurality of battery charging data. Here, the feature map may be an array of numerical data representing features extracted from the corresponding battery charging data, and the numerical value of each feature map may represent how strongly the extracted feature appears in a specific portion of the array. Such a feature map may be visualized in the form of a two-dimensional image, such as a heatmap, which uses color to represent the size or density of values. That is, each pixel value of the image generated in response to the feature map may reflect the data value of the feature map at the corresponding location. Hereinafter, for convenience, the image itself visualizing the feature map will also be referred to as a "feature map," and a more specific process by which the electronic device (100) generates the feature map will be described in detail with reference to FIG. 3 below.

[0059] In step S230, the electronic device (100) may perform image processing on each of the plurality of feature maps based on an image pooling technique. The image pooling technique may be an image processing technique that can reduce the size of a feature map and further emphasize important information about the characteristics of data appearing in the feature map. The image pooling technique may include, for example, a max pooling technique that selects a maximum value corresponding to a plurality of regions included in a target feature map and generates a feature map composed of the maximum values ​​of the plurality of regions, and a dilation pooling technique that generates a feature map composed of representative values ​​of at least some values ​​included in each of the plurality of regions.

[0060] Specifically, the max pooling technique may be a technique for selecting the maximum value from each of a plurality of regions of a preset size included in a target feature map and generating a reduced feature map composed of the selected plurality of maximum values, whereby the most prominent feature may be emphasized and noise may be reduced. Meanwhile, the dilated pooling technique may be a technique for selecting at least some values ​​within each region at intervals equal to a preset dilation rate, deriving a representative value such as a maximum value or an average value from at least some of the selected values ​​for each region, and generating a reduced feature map composed of the derived plurality of representative values, whereby spatial information where information about the characteristics of the data is located may be more efficiently reflected. According to an embodiment, the max pooling technique or the dilated pooling technique may be performed based on the characteristics of the plurality of feature maps, and in addition, various pooling techniques may be performed that can further simplify and clarify the characteristics of the target feature map depending on the purpose of image processing and the characteristics of the target feature map.

[0061] By performing image processing on each of the multiple feature maps in this manner, missing values ​​and outliers can be removed from each feature map, thereby further emphasizing each feature's characteristics. In other words, image processing can improve the distinguishability between feature maps, thereby facilitating smoother clustering.

[0062] In step S240, the electronic device (100) can calculate the similarity between each of the plurality of feature maps on which image processing has been performed. Here, the similarity between each of the plurality of feature maps may be the similarity between images. The similarity between the feature map images may be calculated based on, for example, a structural similarity index measure (SSIM) that measures similarity based on visual similarity. Specifically, SSIM is an index that compares images by measuring differences between structural characteristics of images, such as brightness, contrast, and structural patterns, and can reflect visual characteristics perceived by humans beyond simple differences between pixel values.

[0063] In step S250, the electronic device (100) can determine the optimal number of clusters for the plurality of feature maps on which image processing has been performed based on the calculated similarity, the clustering algorithm, and the index for evaluating the clustering result. Here, the clustering algorithm may be an algorithm that divides the data to be performed into clusters with similar characteristics (e.g., the calculated similarity between images), and the index may be an index that quantitatively evaluates the performance of clustering based on the cohesion of each cluster according to the clustering (a measure of how closely data within the same cluster are grouped) and the separation between clusters (a measure of how clearly different clusters are distinguished) (separation). Meanwhile, the optimal number of clusters may mean the number of clusters that can cluster the plurality of feature maps more efficiently and with distinct boundaries. A more specific embodiment in which the electronic device (100) determines the optimal number of clusters for multiple feature maps based on similarity, clustering algorithm, and index will be described in detail with reference to FIG. 12 below.

[0064] At step S260, the electronic device (100) can cluster multiple feature maps subjected to image processing into clusters with an optimal number of clusters. By performing this clustering, battery charging patterns can be further refined, and types that enable objective classification can be identified.

[0065] FIG. 3 illustrates a process for generating a feature map based on data regarding a battery charging pattern according to one embodiment.

[0066] According to one embodiment, the electronic device (100) may obtain a plurality of charging data for a plurality of vehicles (step S310), and may identify one or more charging sections based on first battery charging data for a first vehicle among the plurality of battery charging data for the plurality of vehicles (step S320). Subsequently, the electronic device (100) may identify the SoC of the battery at the start time of charging and the SoC of the battery at the end time of charging for each of the one or more charging sections (step S330), and may generate a first feature map indicating the frequency of one or more charging sections corresponding to the range of the SoC of the battery at the start time of charging and the range of the SoC of the battery at the end time of charging (step S340). That is, the electronic device (100) may generate the first feature map as a feature map regarding the charging pattern of the first battery charging data for the first vehicle, and may generate each feature map for each of the plurality of vehicles.

[0067] More specifically, the electronic device (100) may analyze the frequency of charging sections based on the first battery charging data and estimate a probability density function regarding the frequency of each charging section. Subsequently, the electronic device (100) may generate a feature map in the form of an image representing the frequency of the battery charging sections of the corresponding vehicle based on the estimated probability density function. By performing this process for each of a plurality of battery charging data, the electronic device (100) may generate a plurality of feature maps regarding battery charging patterns.

[0068] Each of the multiple feature maps in image form can represent a specific battery charging pattern extracted from battery charging data for a specific vehicle. Specifically, if the users of the vehicles for which battery charging data is collected are different for each vehicle, each feature can represent a charging pattern for each user of the corresponding vehicle.

[0069] Figure 4 shows data regarding a battery charging pattern according to one embodiment.

[0070] Referring to FIG. 4, the first battery charging data for the first vehicle may be in the form of a graph (400) with time as the x-axis and SoC as the y-axis. Specifically, the graph (400) may include SoC information of the first battery of the first vehicle for a preset period of time, and more specifically, may include information regarding a charging section confirmed each time the first battery is charged. Accordingly, the graph (300) may indicate at what SoC level charging starts and ends each time the first battery is charged.

[0071] Figure 5 shows statistical data regarding a charging section according to one embodiment.

[0072] Referring to FIG. 5, the electronic device (100) may generate a statistical data graph (500) that analyzes the frequency of the charging section of the first battery based on the first battery charging data graph (400). The statistical data graph (500) may comprehensively represent the charging start SoC, the charging end SoC, and the degree to which each charging section is repeated, and may be in the form of a graph with the charging start SoC as the x-axis and the charging end SoC as the y-axis. The frequency of each charging section may be represented by a color shade or pattern corresponding to each charging section, and in the graph (500), the frequency is represented by the density of the pattern corresponding to the cell representing each charging section. That is, each cell having a pattern in the graph (500) indicates that there was a charging action corresponding to the charging section from the SoC at which charging started to the SoC at which charging ended, and a form having a lower density of a pattern may indicate a higher frequency of charging in the corresponding charging section. Accordingly, the statistical data graph (500) may indicate that the charging frequency is relatively high in the section where charging starts at about 20-30% of SoC and ends at about 10-20% of SoC, and in the section where charging starts at about 20-30% of SoC and ends at about 70-80% of SoC. Meanwhile, by generating the statistical data graph (500) in this way, charging sections and meaningless data, etc., may be refined among the first battery charging data.

[0073] Figure 6 shows a probability distribution graph regarding the frequency of charging intervals according to one embodiment.

[0074] Referring to FIG. 6, the electronic device (100) can estimate a probability density function for the frequency of each charging section based on the statistical data graph (500). As a more specific example, the electronic device (100) can estimate the probability density function for each charging section confirmed based on the statistical data graph (500) using a Kernel Density Estimation (KDE) technique that estimates the probability density function through a kernel function. Then, based on the estimated probability density function, the electronic device (100) can generate a graph (600) that visualizes the probability distribution of the charging section. Thus, the estimated probability distribution graph (600) can represent the estimated probability density for each charging section in the form of a three-dimensional graph with the charging start SoC as the x-axis, the charging end SoC as the y-axis, and the probability for the corresponding SoC section as the z-axis. Meanwhile, although not shown in FIG. 6, depending on the embodiment, a graph may be generated to indicate the size of the probability or the frequency of the charging section through shading or patterns of the probability distribution graph.

[0075] Figure 7 illustrates a feature map according to one embodiment.

[0076] Referring to FIG. 7, the electronic device (100) may generate a first feature map based on a probability distribution graph (600) of a confirmed charging section for the first battery, and may visualize the first feature map as a feature map (700) in the form of a two-dimensional heat map image with the charging start SoC as the x-axis and the charging end SoC as the y-axis. Here, the shading of the feature map (700) may indicate the frequency of the charging section. That is, the darker the shading in the feature map (700), the more frequently the first battery is charged. Accordingly, each feature map may indicate a charging habit or tendency regarding the SoC at which charging mainly starts and ends in the corresponding vehicle, and as a more specific example, the feature map (700) may indicate that the user of the first vehicle has a charging habit of mainly starting charging at a low SoC and ending charging at a low SoC.

[0077] Figure 8 illustrates an image processing process based on a maximum pooling technique according to one embodiment.

[0078] Referring to FIG. 8, the electronic device (100) can perform image processing on a feature map (810) based on a maximum pooling technique. To perform the pooling technique, the electronic device (100) can visualize the feature map based on a pixel coordinate system. That is, the feature map (810) can be in the form of an image in which the x-axis coordinate of a pixel is the x-axis and the y-axis coordinate of a pixel is the y-axis, and the feature map (820) on which maximum pooling is performed can be in the form of an image that is reduced in size compared to the feature map (810). The feature map (820) can more clearly represent the features of the image according to maximum pooling.

[0079] Figure 9 illustrates an image processing process based on an extended pooling technique according to one embodiment.

[0080] The description in FIG. 8 will be equally applicable to FIG. 9, except that the extended pooling technique is performed instead of the max pooling technique. The feature map (910) on which extended pooling is performed is an image format that is reduced in size compared to the feature map (810), and the features of the image can be more clearly expressed according to the extended pooling.

[0081] Figures 10a to 10c illustrate multiple feature maps on which a pooling technique is performed according to one embodiment.

[0082] Figures 10a to 10c illustrate the effect of pooling techniques, which simplify images and sharpen their features, transforming previously disparate feature maps into identical or very similar feature maps. In each figure, a class refers to an arbitrary group classified as an image of the same or very similar shape, and multiple feature maps classified into each class and multiple feature maps on which pooling has been performed are illustrated.

[0083] Referring to FIG. 10A, class 0 (1010) may include a plurality of feature maps (1011), and each of the plurality of feature maps (1011) may include a plurality of feature maps (1012) on which pooling has been performed. Referring to the plurality of feature maps (1012) on which pooling has been performed, the features of each feature map, which are indicated by shades, are simplified due to pooling, making it easier to determine that they belong to the same group, and thereby, the electronic device (100) may efficiently and quickly perform clustering on the feature maps on which pooling has been performed. FIG. 10b illustrates a plurality of feature maps (1021) belonging to class 1 (1020) and a plurality of feature maps (1022) on which pooling has been performed, and FIG. 10c illustrates a plurality of feature maps (1031) belonging to class 2 (1030) and a plurality of feature maps (1032) on which pooling has been performed, respectively, and the description with respect to FIG. 10a can be equally applied.

[0084] Figure 11 illustrates a clustering process based on similarity between images according to one embodiment.

[0085] According to one embodiment, the electronic device (100) may obtain a plurality of battery charging data for a plurality of vehicles (step S1110). Subsequently, the electronic device (100) may generate a plurality of feature maps regarding battery charging patterns based on each of the plurality of battery charging data (step S1120), and may perform image processing on each of the generated plurality of feature maps based on an image pooling technique (step S1130).

[0086] The electronic device (100) can then calculate the similarity between each of the plurality of feature maps on which image processing has been performed based on the structural similarity index for clustering (step S1140). That is, the electronic device (100) can measure the structural similarity between each of the plurality of feature maps whose features have been simplified through image processing, thereby calculating a similarity that effectively reflects the visual similarity between each feature map, and can utilize this similarity as the basis for performing clustering.

[0087] The electronic device (100) can determine the optimal number of clusters for a plurality of feature maps on which image processing has been performed based on the similarity calculated as above, the clustering algorithm, and the index for evaluating the clustering result (step S1150), and can perform clustering of the plurality of feature maps on which image processing has been performed into clusters with the optimal number of clusters (step S1160).

[0088] Figure 12 illustrates a clustering process based on a clustering algorithm and index according to one embodiment.

[0089] According to one embodiment, the electronic device (100) may obtain a plurality of battery charging data for a plurality of vehicles (step S1210), and may generate a plurality of feature maps regarding battery charging patterns based on each of the plurality of battery charging data (step S1220). Subsequently, the electronic device (100) may perform image processing on each of the plurality of generated feature maps based on an image pooling technique (step S1230), and may calculate a similarity between each of the plurality of feature maps on which image processing has been performed (step S1240).

[0090] The electronic device (100) can then perform clustering multiple times on the multiple feature maps on which image processing has been performed so that the multiple feature maps have different numbers of clusters according to the calculated similarity and the clustering algorithm (step S1250). Specifically, the electronic device (100) can perform clustering multiple times on the multiple feature maps on which image processing has been performed so that the multiple feature maps on which image processing has been performed have k (k∈{2, 3, ..., k}) clusters according to the clustering algorithm. That is, the electronic device (100) can cluster the multiple feature maps on which image processing has been performed so that the number of clusters varies each time clustering is performed.

[0091] In one embodiment, the clustering algorithm performed on the plurality of feature maps on which image processing has been performed may be any type of clustering algorithm that can set the number of clusters as a parameter (i.e., be hyperparameterized). For example, it may be a hierarchical clustering algorithm that hierarchizes at least some features included in each data set based on similarity. The hierarchical clustering algorithm may have the advantage of performing very little computation because it repeatedly generates an initial similarity matrix based on similarity between images, such as a structural similarity index, and then merges or divides clusters with high similarity into sub-clusters based on the matrix.

[0092] Next, the electronic device (100) can determine the optimal number of clusters for the plurality of feature maps on which image processing has been performed based on the clustering results performed multiple times and the indexes for evaluating the clustering results (step S1260). Here, the indexes may include, for example, a first index, a second index, and a third index, and the electronic device (100) can calculate the first index according to the number of clusters, the second index according to the number of clusters, and the third index according to the number of clusters, respectively, based on the clustering results performed multiple times. The index according to the number of clusters may refer to a performance index that evaluates the results of performing clustering with different k values, and may be organized in the form of a graph, for example. Meanwhile, the types of indexes that the electronic device (100) calculates based on the clustering results performed multiple times are not limited to the three types of indexes, and it will be clearly understood by those skilled in the art that more types of indexes may be calculated depending on the embodiment.

[0093] According to one embodiment, each of the first index, the second index, and the third index may include at least one of a Silhouette Index technique, a Calinski-Harabasz Index technique, and a Davies-Bouldin Index technique, which evaluate clustering results based on the cohesion of each cluster and the separation between clusters, for each clustering result performed according to a clustering algorithm. The techniques have differences in the process of calculating cohesion and separation. For example, according to the Silhouette Index technique, cohesion may be calculated based on the average of the SSIM between each data point and all other data points in the same cluster, and separation may be calculated based on the average of the SSIM between each cluster and all data points in the closest other cluster. On the other hand, according to the Kalinsky-Harabats index technique, cohesion is calculated based on the sum of the SSIM between all data points in each cluster and its centroid, and separability can be calculated based on the sum of the SSIM between the centroid of each cluster and all data points in other clusters. Furthermore, according to the Davis-Boldin index technique, cohesion is calculated based on the average of the SSIM between each data point in each cluster and its centroid, and separability can be calculated based on the average of the SSIM between all data points in each cluster and all data points in other clusters.

[0094] Next, the electronic device (100) may determine an optimal number of clusters for the first data set based on a first cluster number corresponding to a maximum or minimum value among the first indexes according to the calculated number of clusters, a second cluster number corresponding to a maximum or minimum value among the second indexes according to the calculated number of clusters, and a third cluster number corresponding to a maximum or minimum value among the third indexes according to the calculated number of clusters. Here, each of the maximum or minimum value among the first indexes, the maximum or minimum value among the second indexes, and the maximum or minimum value among the third indexes may correspond to an optimal value among the maximum and minimum values ​​set based on the characteristics of the indexes. For example, when the index is a silhouette index technique or a Kalinsky-Harabats index technique, the maximum value may correspond to the optimal value, and when the index is a Davis-Boldin index technique, the minimum value may correspond to the optimal value. Meanwhile, the optimal value for each index is not limited to the maximum or minimum value, and can be set to an appropriate value depending on the characteristics of the index.

[0095] According to one embodiment, the optimal number of clusters for a plurality of feature maps on which image processing has been performed may be set to an overlapping value among the first number of clusters, the second number of clusters, and the third number of clusters. That is, the number of clusters where the maximum or minimum values ​​intersect as evaluated by each of the three indices may be set as the optimal number of clusters. For example, if the first number of clusters and the second number of clusters are 9, and the third number of clusters is 8, the optimal number of clusters may be set to 9. In this way, by setting the number of clusters evaluated as having the optimal value according to various index techniques as the optimal number of clusters for the first data set, clustering can be performed with guaranteed high performance. According to another embodiment, if the first number of clusters, the second number of clusters, and the third number of clusters all have different values, their average value may be set as the optimal number of clusters for a plurality of feature maps on which image processing has been performed.

[0096] Next, the electronic device (100) can perform clustering of multiple feature maps on which image processing has been performed into clusters with an optimal number of clusters (step S1270). That is, multiple feature maps on which image processing has been performed can be clustered with an optimal number of clusters set as described above.

[0097] Figure 13 illustrates a process for determining the optimal number of clusters according to one embodiment.

[0098] Referring to FIG. 13, which illustrates a process (1300) for determining an optimal number of clusters for a plurality of feature maps (1310) on which multiple image processing has been performed, the electronic device (100) can perform multiple clustering with k cluster numbers (number of clusters) for a plurality of feature maps (1310) on which image processing has been performed. As described above, the multiple clustering can be performed with different cluster numbers for different k values. Subsequently, the electronic device (100) can calculate a first index (1321) according to the k value, a second index (1322) according to the k value, and a third index (1323) according to the k value, and can determine an optimal number of clusters for a plurality of feature maps (1310) on which image processing has been performed based on the first index (1321), the second index (1322) according to the k value, and the third index (1323) according to the k value.

[0099] Figure 14a shows a first index according to the number of clusters according to one embodiment.

[0100] Here, the first index may be a silhouette index. The graph (1400) represents the silhouette index evaluated for each value of the number of clusters k ranging from 2 to 10, with the number of clusters (k) on the x-axis and the silhouette index on the y-axis. In the case of the silhouette index technique, the maximum value is the optimal value, so the number of clusters corresponding to the optimal value among the first indices may be 8.

[0101] Figure 14b shows a second index according to the number of clusters, according to one embodiment.

[0102] Here, the second index may be the Kalinsky-Harabatz index. Graph (1410) represents the Kalinsky-Harabatz index (CHI) evaluated for each value of the number of clusters k ranging from 2 to 10, with the number of clusters (k) on the x-axis and the Kalinsky-Harabatz index on the y-axis. Considering that the maximum value is the optimal value in the Kalinsky-Harabatz index technique, the number of clusters corresponding to the optimal value among the second indices may be 8, as in Fig. 14a.

[0103] Figure 14c shows a third index according to the number of clusters, according to one embodiment.

[0104] Here, the third index may be the Davis-Boldin index. Graph (1420) represents the Davis-Boldin index (DBI) evaluated for each value of the number of clusters k ranging from 2 to 10, with the number of clusters (k) on the x-axis and the Davis-Boldin index on the y-axis. Considering that the minimum value is the optimal value in the Davis-Boldin index technique, the number of clusters corresponding to the optimal value among the third indices may be 8, as in FIGS. 14a and 14b.

[0105] That is, in the examples of FIGS. 14a to 14c, the optimal number of clusters of the target data set determined based on the produced indices may be 8.

[0106]

[0107] The electronic device according to the above-described embodiments may include a processor, a memory for storing and executing program data, permanent storage such as a disk drive, a communication port for communicating with an external device, a user interface device such as a touch panel, a key, a button, etc. The methods implemented as software modules or algorithms may be stored on a computer-readable recording medium as computer-readable codes or program instructions executable on the processor. Here, the computer-readable recording medium includes a magnetic storage medium (e.g., read-only memory (ROM), random-access memory (RAM), floppy disk, hard disk, etc.) and an optical reading medium (e.g., CD-ROM, DVD: Digital Versatile Disc)). The computer-readable recording medium may be distributed to computer systems connected to a network, so that the computer-readable code may be stored and executed in a distributed manner. The medium may be readable by a computer, stored in a memory, and executed by a processor.

[0108] The present embodiment may be represented by functional block configurations and various processing steps. These functional blocks may be implemented by various hardware and / or software configurations that perform specific functions. For example, the embodiment may employ direct circuit configurations such as memory, processing, logic, look-up tables, etc., which may perform various functions under the control of one or more microprocessors or other control devices. Similarly, the present embodiment may be implemented in a programming or scripting language such as C, C++, Java, assembler, etc., including various algorithms implemented as a combination of data structures, processes, routines, or other programming configurations. Functional aspects may be implemented as algorithms that execute on one or more processors. Furthermore, the present embodiment may employ conventional techniques for electronic configuration, signal processing, and / or data processing. Terms like "mechanism," "element," "means," and "composition" can be used broadly and are not limited to mechanical or physical components. These terms can also encompass a series of software routines, such as those associated with a processor.

[0109] The above-described embodiments are merely examples, and other embodiments may be implemented within the scope of the claims set forth below.

Claims

1. In electronic devices, transceiver; processor; and Contains one or more memories that store one or more instructions, The one or more instructions, when executed, cause the processor to: Obtain multiple battery charging data for multiple vehicles, Based on each of the plurality of battery charging data, a plurality of feature maps regarding battery charging patterns are generated, Based on the image pooling technique, image processing is performed on each of the plurality of feature maps, Compute the similarity between each of the plurality of feature maps on which the above image processing has been performed, Based on the calculated similarity, clustering algorithm, and index for evaluating the clustering result, the optimal number of clusters for the multiple feature maps on which the image processing was performed is determined, An electronic device configured to perform clustering of a plurality of feature maps on which the above image processing has been performed into clusters having the optimal number of clusters.

2. In paragraph 1, The above multiple battery charging data is, An electronic device comprising, for each of a plurality of batteries included in the plurality of vehicles, data collected over a preset period of time, information on the state of charge (SoC) of the battery over time.

3. In paragraph 2, The one or more instructions, when executed, cause the processor to: Based on the first battery charging data for the first vehicle among the plurality of battery charging data for the plurality of vehicles, one or more charging sections are identified, For each of the above one or more charging sections, check the SoC of the battery at the start of charging and the SoC of the battery at the end of charging, An electronic device configured to generate a first feature map indicating a frequency of one or more charging intervals corresponding to a range of SoC of the battery at a charging start time and a range of SoC of the battery at a charging end time.

4. In paragraph 3, An electronic device wherein the frequency of one or more of the charging intervals is determined based on a probability density function.

5. In paragraph 1, The above image pooling technique is, An electronic device comprising a max pooling technique for selecting a maximum value corresponding to a plurality of regions included in a target feature map and generating a feature map composed of the maximum values ​​of the plurality of regions, and a dilation pooling technique for generating a feature map composed of representative values ​​of at least some values ​​included in each of the plurality of regions.

6. In paragraph 1, The one or more instructions, when executed, cause the processor to: An electronic device configured to calculate the similarity between each of a plurality of feature maps on which image processing is performed based on a structural similarity index.

7. In paragraph 1, The one or more instructions, when executed, cause the processor to: For the plurality of feature maps on which the above image processing has been performed, clustering is performed multiple times to have different numbers of clusters according to the calculated similarity and the clustering algorithm, An electronic device configured to determine the optimal number of clusters for a plurality of feature maps on which image processing has been performed based on the results of the clustering performed multiple times and an index for evaluating the results of the clustering.

8. In paragraph 7, The above clustering algorithm is, An electronic device comprising a hierarchical clustering algorithm that hierarchizes at least some feature maps included in a plurality of feature maps on which image processing is performed based on similarity.

9. In paragraph 7, The above index includes a first index, a second index, and a third index, The one or more instructions, when executed, cause the processor to: Based on the results of the clustering performed multiple times, the first index according to the number of clusters, the second index according to the number of clusters, and the third index according to the number of clusters are each calculated, An electronic device configured to determine an optimal number of clusters for a plurality of feature maps on which image processing has been performed based on a first number of clusters corresponding to a maximum or minimum value among the first indexes according to the calculated number of clusters, a second number of clusters corresponding to a maximum or minimum value among the second indexes according to the calculated number of clusters, and a third number of clusters corresponding to a maximum or minimum value among the third indexes according to the calculated number of clusters.

10. In paragraph 9, Each of the first index, the second index and the third index, An electronic device comprising at least one of the Silhouette Index technique, the Calinski-Harabasz Index technique, and the Davies-Bouldin Index technique, which evaluates the clustering results based on the cohesion of each cluster and the separation between clusters, for each of the calculated similarity and the clustering results performed according to the clustering algorithm.

11. A method for clustering data regarding battery charging patterns performed by an electronic device, A step of acquiring multiple battery charging data for multiple vehicles; A step of generating a plurality of feature maps regarding a battery charging pattern based on each of the plurality of battery charging data; A step of performing image processing on each of the plurality of feature maps based on an image pooling technique; A step of calculating the similarity between each of a plurality of feature maps on which the above image processing has been performed; A step of determining the optimal number of clusters for a plurality of feature maps on which image processing has been performed based on the calculated similarity, clustering algorithm, and index for evaluating the clustering result; and A method for clustering data regarding a battery charging pattern, comprising a step of clustering a plurality of feature maps on which the above image processing has been performed into clusters having the optimal number of clusters.

12. A non-transitory computer-readable recording medium having recorded thereon a program for executing the method of Article 11 on a computer.

Images