Data analysis apparatus and operating method thereof

Abstract

According to an embodiment disclosed in this document, a data analysis device may comprise: an information acquisition unit for acquiring a plurality of data sets, each of which contains time-series data of a plurality of parameters; and a controller that extracts a feature of each of the plurality of data sets based on a correlation between the plurality of parameters, extracts a pattern of each of the plurality of data sets based on a similarity between the extracted features, and classifies the plurality of data sets into a plurality of clusters based on the pattern.

Description

TECHNICAL FIELD

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This application claims priority to Korean Patent Application No. 10-2023-0106325, filed on August 14, 2023, the entire contents of which are incorporated herein by reference for all purposes. TECHNICAL FIELD

[0003] Embodiments disclosed in the present document relate to a data analysis device and an operating method thereof. BACKGROUND

[0004] In recent years, research and development of secondary batteries have been actively conducted. Here, the secondary battery is a rechargeable battery, and includes all conventional Ni / Cd batteries, Ni / MH batteries, and recent lithium ion batteries. Among these secondary batteries, the lithium ion battery has the advantage of having much higher energy density than the conventional Ni / Cd battery and the Ni / MH battery. In addition, the lithium ion battery can be manufactured to be small and light enough to be used as a power source for mobile devices, and recently, its use range has been expanded to a power source for electric vehicles, attracting attention as a next-generation energy storage medium.

[0005] Generally, a battery is monitored and controlled by a battery management system (BMS). Data collected from the battery management system includes data on various factors, and the amount of data is so large that it is difficult to analyze the cause of an anomaly even if an anomaly occurs in the battery. SUMMARY

[0006] TECHNICAL PROBLEM

[0007] An object of embodiments disclosed in the present document is to provide a data analysis device capable of effectively identifying the cause of an anomaly in a battery by classifying battery data, and an operating method thereof.

[0008] The technical objects of embodiments disclosed in the present document are not limited to the above-mentioned objects, and other objects not described herein will be clearly understood by persons skilled in the art from the following description.

[0009] TECHNICAL SOLUTION

[0010] The data analysis device according to embodiments disclosed in the present document can include an information acquisition unit configured to acquire a plurality of data sets each containing time series data of a plurality of parameters, and a controller configured to extract a feature of each data set based on a correlation between the plurality of parameters, extract a pattern of each data set based on a similarity between the extracted features, and classify the plurality of data sets into a plurality of clusters based on the pattern.

[0011] According to an embodiment, the controller can extract the features as a matrix representing a correlation between the plurality of parameters.

[0012] According to an embodiment, the correlation between the plurality of parameters can be represented by a Pearson correlation coefficient of each pair of parameters.

[0013] According to an embodiment, the controller can calculate a correlation coefficient between the features of the plurality of data sets, and can generate a correlation coefficient matrix composed of the correlation coefficients.

[0014] According to an embodiment, the controller can apply a classification algorithm to the correlation coefficient matrix to extract the pattern.

[0015] According to an embodiment, the classification algorithm can be a K-means algorithm.

[0016] According to an embodiment, the controller can extract the pattern based on a result of applying the classification algorithm to the correlation coefficient matrix as the value of K is changed.

[0017] According to an embodiment, the controller can analyze a characteristic of each of the plurality of clusters.

[0018] According to an embodiment, the controller can analyze an abnormal occurrence history related to the data set included in each of the plurality of clusters to estimate a parameter associated with the abnormal occurrence history.

[0019] According to an embodiment, the controller can extract the features by setting a window interval for each of the plurality of data sets.

[0020] According to an embodiment disclosed in the present document, an operation method of a data analysis device can include the steps of: acquiring a plurality of data sets each containing time series data of a plurality of parameters; extracting features of each data set based on a correlation between the plurality of parameters; extracting a pattern of each data set based on similarity between the extracted features; and classifying the plurality of data sets into a plurality of clusters based on the pattern.

[0021] According to an embodiment, the step of extracting the features of each data set can include the step of extracting the features as a matrix representing a correlation between the plurality of parameters.

[0022] According to an embodiment, the step of extracting the pattern of each data set can include the steps of: calculating a correlation coefficient between the features of the plurality of data sets, and generating a correlation coefficient matrix composed of the correlation coefficients; and applying a classification algorithm to the correlation coefficient matrix to extract the pattern.

[0023] According to the implementation method, the classification algorithm may be the K-means algorithm, and the step of extracting patterns by applying the classification algorithm may include the following steps: extracting patterns based on the results of applying the classification algorithm to the correlation coefficient matrix when the K value is changed.

[0024] According to the implementation method, the operation method of the data analysis device may further include the following steps: estimating parameters associated with the anomaly occurrence history by analyzing the anomaly occurrence history associated with the datasets included in each of the multiple clusters.

[0025] Beneficial effects

[0026] The data analysis apparatus and its operating method according to the embodiments disclosed in this document are advantageous in facilitating the analysis of large amounts of unstructured data by extracting patterns from datasets and classifying those patterns.

[0027] The data analysis apparatus and its operating method according to the embodiments disclosed in this document are also advantageous in providing a basis for identifying the causes of anomalies by classifying similar datasets.

[0028] In addition, various effects that can be directly or indirectly determined through this document may be provided. Attached Figure Description

[0029] Figure 1 This is a diagram illustrating the configuration of a data analysis system according to an embodiment disclosed in this document.

[0030] Figure 2 This is a block diagram illustrating the configuration of a data analysis device according to an embodiment disclosed in this document.

[0031] Figure 3 This is a block diagram illustrating the configuration of a controller according to an embodiment disclosed in this document.

[0032] Figure 4 This is a diagram illustrating an example of the extracted features according to the embodiments disclosed in this document.

[0033] Figure 5 This is a diagram illustrating an example of a correlation coefficient matrix according to an embodiment disclosed in this document.

[0034] Figure 6 This is a diagram illustrating an example of an extracted pattern according to an embodiment disclosed in this document.

[0035] Figure 7 This is a flowchart illustrating an operation method of a data analysis device according to an embodiment disclosed in this document.

[0036] Figure 8 This is a flowchart illustrating the process of extracting a mode according to the implementation method disclosed in this document.

[0037] Figure 9 This is a block diagram illustrating the hardware configuration of a computing system for performing an operation method of a data analysis device according to an embodiment disclosed in this document. Detailed Implementation

[0038] In the following description, various embodiments of the invention will be described with reference to the accompanying drawings. However, the description is not intended to limit the invention to the specific embodiments described herein, and should be construed as including various modifications, equivalents, and / or alternatives to the embodiments described herein.

[0039] In this document, unless the context explicitly indicates otherwise, the singular noun corresponding to an item may include one or more instances of that item. In this document, phrases such as “A or B,” “at least one of A and B,” “at least one of A or B,” “A, B, or C,” “at least one of A, B, and C,” and “at least one of A, B, or C” may each refer to any single item listed or any combination of these items. Terms such as “first,” “second,” “first,” and “second” are simply used to distinguish one component from others and do not otherwise limit these components (e.g., importance or order). In this document, when a component (e.g., the first component) is referred to as “connected” or “linked” to another component (e.g., the second component) with or without the use of terms such as “functionally” or “communically,” it means that the first component can be connected to the second component directly (e.g., via a wired connection), wirelessly, or through a third component.

[0040] Each component (e.g., module or program) described in this document may include one or more instances. Depending on various implementations, one or more components or operations described herein may be omitted, or one or more other components or operations may be added. Alternatively or additionally, multiple components (e.g., modules or programs) may be integrated into a single component. In this case, the integrated component may perform one or more functions of the corresponding components in the multiple components in the same or similar manner as the functions performed by each of the multiple components prior to integration. Depending on various implementations, operations performed by modules, programs, or other components may be performed sequentially, in parallel, iteratively, or heuristically, or one or more operations may be performed in a different order, omitted, or one or more other operations may be added.

[0041] The terms “module” or “unit” as used in this document can include units implemented in hardware, software, or firmware, and are used interchangeably with terms such as logic, logic block, component, or circuit. A module can be a component assembled as a whole, or it can be one or more units that perform one or more functions, which can constitute the smallest unit of a component or a part thereof. For example, according to one implementation, a module can be implemented as an application-specific integrated circuit (ASIC).

[0042] Various embodiments of this document can be implemented as software (e.g., a program or application) including one or more instructions stored on a machine-readable storage medium (e.g., memory). For example, a processor of the device can invoke and execute at least one instruction stored on the storage medium. This enables the device to operate according to at least one invoked instruction to perform at least one function. The one or more instructions may include code generated by a compiler or code executable by an interpreter. The machine-readable storage medium may be provided in the form of a non-transitory storage medium. Here, "non-transitory" means only that the storage medium is a tangible device, excluding signals (e.g., electromagnetic waves), and the term is used regardless of whether the data is permanently or temporarily stored.

[0043] Figure 1 This is a diagram illustrating the configuration of a data analysis system according to an embodiment disclosed in this document.

[0044] Reference Figure 1 The data analysis system may include battery management systems (BMS) 10, 20, and 30 and a data analysis device 100. Although Figure 3 The document describes three battery management systems, but the number is not limited to three, and battery management systems 10, 20, and 30 can be configured to include n systems (where n is a natural number). In a configuration where the battery pack is mounted on or connected to a vehicle or charger, each battery management system can be located inside or outside the battery pack.

[0045] Data analysis device 100 can acquire datasets of time-series data from each of battery management systems 10, 20, and 30. Data analysis device 100 can classify the acquired datasets into multiple clusters. More specifically, data analysis device 100 can extract features from each of the multiple datasets and classify the datasets into multiple clusters by extracting patterns from the features extracted from each dataset.

[0046] The data analysis device 100 can extract patterns from datasets acquired in real time from each battery management system and classify the datasets. In an implementation, the data analysis device 100 can learn the process of extracting features from the dataset, extracting patterns from the features, and classifying the dataset based on the patterns. When a new dataset is acquired, the data analysis device 100 can use the trained model to classify the new dataset.

[0047] The data analysis device 100 can also learn the relationship between battery data and battery anomalies by using the classification results of the dataset. For example, the data analysis device 100 can classify the dataset and analyze the anomaly history within the classified clusters to identify the causes of the anomalies.

[0048] Through this process, the data analysis device 100 can facilitate the analysis of large amounts of unstructured data by extracting patterns from the dataset and classifying these patterns. Furthermore, the data analysis device 100 can classify similar datasets, thereby providing a basis for identifying the causes of anomalies.

[0049] Figure 2 This is a block diagram illustrating the configuration of a data analysis device according to an embodiment disclosed in this document.

[0050] Reference Figure 2 The data analysis device 100 may include an information acquisition unit 110 and a controller 120.

[0051] The information acquisition unit 110 may include a communication circuit capable of establishing wired and / or wireless communication channels between components (such as servers, vehicles, and battery management systems) inside and outside the data analysis device 100, and can send and receive various types of data through the established communication channels.

[0052] The information acquisition unit 110 can acquire multiple datasets. Each of the multiple datasets may include time-series data with multiple parameters. In an embodiment, the information acquisition unit 110 may acquire multiple datasets from multiple battery management systems. For example, the information acquisition unit 110 may acquire each dataset from different battery management systems.

[0053] Multiple parameters can include a variety of parameters related to battery state. For example, multiple parameters can comprehensively include parameters related to battery state (e.g., current, voltage, temperature), parameters derived from battery state (e.g., SOC, SOH, DOD, voltage deviation), vehicle-related parameters when the battery is installed in the vehicle (e.g., driving time, driving distance, speed, acceleration), and parameters related to external factors that may affect battery state (e.g., ambient temperature), and are not limited to the examples above.

[0054] The controller 120 can extract features from each of multiple datasets based on the correlation between multiple parameters. For example, the controller 120 can extract features from each of multiple datasets based on various operations (e.g., covariance, joint probability, etc.) that can calculate the correlation between multiple parameters included in the time series data of each dataset.

[0055] According to the implementation, the controller 120 can extract features in matrix form representing the correlation between multiple parameters. For example, the features of each of the multiple datasets can be represented as a matrix representing the correlation. For example, when the dataset obtained by the information acquisition unit 110 includes time series data with h parameters (where h is a natural number), the features extracted by the controller 120 can be represented as h... The h matrix can be used to calculate the correlation between multiple parameters based on various operations (e.g., covariance, joint probability, etc.) from data of any two parameters.

[0056] According to the implementation method, the correlation between multiple parameters can be expressed as the Pearson correlation coefficient of each parameter pair. For example, when the dataset obtained by the information acquisition unit 110 includes time series data with h parameters (where h is a natural number), the number of parameter pairs can be h². In this case, the features of each dataset are represented as h. The h matrix, where the value of each component is the Pearson correlation coefficient of the parameter pair.

[0057] In the implementation, the Pearson correlation coefficient can be defined as shown in Equation 1.

[0058] [Formula 1]

[0059] Here, r xy This represents the Pearson correlation coefficient between parameters x and y, where n represents the number of data points included in the time series data for each parameter. i Let y represent the i-th data point in the time series data of parameter x. i This represents the i-th data point in the time series data of parameter y. This represents the mean of the data for parameter x, and This represents the mean of the parameter y data.

[0060] In other words, the Pearson correlation coefficient can be derived from the covariance and standard deviation of the two parameters. The Pearson correlation coefficient has a normalized value, and its range, for example, can be between -1 and 1. As an example, the Pearson correlation coefficient between the same parameters is 1; therefore, the correlation coefficient between parameters h is 1. The elements on the diagonal of matrix h can have a value of 1.

[0061] According to the implementation, controller 120 can extract patterns from each of multiple datasets based on the similarity between the extracted features. For example, controller 120 can cluster each of multiple datasets with similar features. Controller 120 can extract patterns from each dataset based on the clustering results.

[0062] In another example, controller 120 can analyze the features of each dataset to transform those features into patterns. As an example, controller 120 can define multiple patterns and determine the pattern for each dataset based on its features. In this case, controller 120 can learn the process of determining patterns from the features of the datasets.

[0063] According to the implementation, controller 120 can calculate the correlation coefficient between features of multiple datasets. Controller 120 can derive the correlation coefficient based on various operations (e.g., covariance, joint probability, etc.). For example, controller 120 can calculate the correlation coefficient between features based on Equation 2 below.

[0064] [Equation 2]

[0065] Here, A and B each represent feature matrices, and r represents the correlation coefficient between the features of A and B. nm B represents the matrix value of the (n, m) element in characteristic matrix A. nm Let (n, m) represent the matrix value of the (n, m) element in the characteristic matrix B. Let represent the mean of the characteristic matrix A, and This represents the mean of the characteristic matrix B.

[0066] According to the implementation method, the controller 120 can generate a correlation coefficient matrix composed of correlation coefficients. For example, when there are i datasets (where i is a natural number), the correlation coefficient matrix can be i... matrix i.

[0067] According to the implementation, the controller 120 can apply a classification algorithm to the correlation coefficient matrix to extract patterns from each of the multiple datasets. The classification algorithm can include various algorithms, such as decision tree-based algorithms (e.g., XGBoost), K-nearest neighbor (K-NN) classification algorithms (e.g., K-means), and support vector machine (SVM) algorithms.

[0068] Depending on the implementation method, the classification algorithm can be the K-means algorithm. The K-means algorithm is an algorithm that classifies multiple data points into K clusters. For example, the K-means algorithm can classify data points based on the centroid of each cluster and the average distance between the data points.

[0069] According to the implementation, the controller 120 can extract patterns based on the results of applying a classification algorithm to the correlation coefficient matrix when the value of K is changed. In the K-means algorithm, the value of K can represent the number of classification clusters. The controller 120 can represent the classification result numerically according to the value of K. For example, when K=2, each dataset can belong to a first cluster or a second cluster. In this case, the controller 120 can assign a classification result value of 0 to the dataset belonging to the first cluster and a classification result value of 1 to the dataset belonging to the second cluster.

[0070] In this way, controller 120 can accumulate classification results as the value of K changes. For example, based on the value of K, controller 120 can determine the classification result value for each dataset as any value from 0 to K-1. Controller 120 can represent the pattern of each dataset as a list of classification result values ​​by accumulating the classification results. As an example, controller 120 can represent the pattern by arranging the classification result values ​​in ascending order based on the value of K.

[0071] According to the implementation, controller 120 can classify multiple datasets into multiple clusters based on patterns. For example, controller 120 can classify multiple datasets into multiple clusters based on the similarity of their patterns. As an example, controller 120 can classify datasets with the same pattern as belonging to the same cluster.

[0072] According to the implementation method, controller 120 can analyze the characteristics of each of multiple clusters. For example, controller 120 can analyze the common characteristics of datasets belonging to each of the multiple clusters.

[0073] According to the implementation, controller 120 can analyze the anomaly occurrence history associated with datasets included in each of the multiple clusters. The anomaly occurrence history can include various battery-related anomalies, such as fire occurrence history, voltage anomaly history, and lithium deposition history.

[0074] The controller 120 can classify the dataset into multiple clusters and analyze the anomaly occurrence history associated with the dataset to estimate the cause of the anomalies. For example, when the same anomaly occurrence history exists within the same cluster, it can be inferred that the cause of the anomaly is similar or the same. The anomaly occurrence history associated with the dataset can include, for example, anomalies occurring in batteries managed by a battery management system from which the dataset is obtained.

[0075] According to the implementation, controller 120 can estimate parameters associated with anomaly occurrence history. Controller 120 can estimate the parameters identified as associated with anomaly occurrence history among a plurality of parameters. Controller 120 can learn the relationship between patterns in the dataset and anomaly occurrence history to estimate the parameters.

[0076] According to one implementation, the controller 120 can set a window interval for each of a plurality of datasets to extract features. Here, each of the plurality of datasets may include time series data, and the time series data may contain multiple time intervals. The controller 120 can perform feature extraction on data corresponding to a specific time frame within the time series data. For example, the controller 120 can set a window interval to cover time intervals with high data reliability and exclude intervals where the time series data is inaccurate (e.g., the initial driving interval).

[0077] Figure 3 This is a block diagram illustrating the configuration of a controller according to an embodiment disclosed in this document.

[0078] Reference Figure 3 The controller 120 may include a feature extraction module 121 and a classification module 123. The controller 120 may use the feature extraction module 121 and the classification module 123 to classify each dataset into multiple clusters. The controller 120 may also include a cause identification module 125.

[0079] The feature extraction module 121 can extract features from each of multiple datasets. The feature extraction module 121 can also generate a correlation coefficient matrix from the extracted features.

[0080] The classification module 123 can extract patterns from each dataset from the extracted features. The classification module 123 can apply a classification algorithm to the features of each dataset to extract patterns. For example, the classification module 123 may include a procedure that performs a K-means algorithm, changes the K value, and performs clustering to extract patterns from each dataset. The classification module 123 can classify multiple datasets based on patterns.

[0081] The cause identification module 125 can analyze the anomaly occurrence history based on the classification results of the dataset and the anomalies present in each cluster to determine the cause of the anomaly. For example, the cause identification module 125 can estimate the parameters identified as being related to the anomaly occurrence history from among multiple parameters.

[0082] According to the implementation, the feature extraction module 121, the classification module 123, and the cause identification module 125 may each include a neural network model. The feature extraction module 121 can learn a process for extracting features from each dataset. The classification module 123 can learn a process for extracting patterns from features and classifying datasets based on those patterns. The cause identification module 125 can learn a process for estimating the causes of anomalies based on the classification results of the datasets and the anomaly occurrence history associated with the datasets included in each cluster.

[0083] Even when a new dataset is acquired, the controller 120 can use the trained feature extraction module 121, classification module 123, and cause identification module 125 to classify the dataset and estimate the cause of the anomaly when an anomaly occurs in the battery associated with the dataset.

[0084] Figure 4 This is a diagram illustrating an example of the extracted features according to the embodiments disclosed in this document.

[0085] Reference Figure 4 The controller 120 can extract features from each dataset. Each feature can be represented as a matrix indicating the correlation between multiple parameters. For example, the value of each element in the feature matrix can be the Pearson correlation coefficient for each pair of parameters.

[0086] exist Figure 4 In the matrix, each row and column represents a parameter (f1 to fn), and the elements of the matrix can be represented as Pearson correlation coefficients between the element and the parameter corresponding to the row and column in which the element is located. In implementations, the elements in the feature matrix can be represented numerically or distinguished by color, such as... Figure 4 As shown.

[0087] Figure 5 This is a diagram illustrating an example of a correlation coefficient matrix according to an embodiment disclosed in this document.

[0088] Reference Figure 5 The controller 120 can generate a correlation coefficient matrix from the features of each dataset. Figure 5 In the diagram, M1 to M4 represent the feature matrices for each dataset, and each element represents the correlation coefficient between features. Figure 5 The quantities and values ​​shown are for illustrative purposes only and are not limited to these examples. Figure 5 As shown, the correlation coefficient for each dataset can be a normalized value between -1 and 1.

[0089] Figure 6 This is a diagram illustrating an example of an extracted pattern according to an embodiment disclosed in this document.

[0090] Reference Figure 6 The controller 120 can extract the pattern for each dataset. A through N represent datasets, and the pattern for each dataset is as follows: Figure 6 The sixth column of the matrix is ​​shown.

[0091] In this implementation, the controller 120 can change the K value to classify the dataset and accumulate classification results. For example, the controller 120 can determine the classification result as one of a value from 0 to K-1 based on the K value. Figure 6In the diagram, columns one through five represent the classification results obtained by setting the K value to 2 through 6, respectively. Controller 120 can generate patterns by listing the classification result values ​​for each dataset. For example, the pattern for dataset A can be defined as 00044, while the pattern for dataset H can be defined as 10112.

[0092] Based on these patterns, controller 120 can classify multiple datasets. For example, since datasets F to J share the same pattern 10112, controller 120 can classify these datasets into the same cluster.

[0093] Controller 120 can analyze the anomaly occurrence history associated with the datasets included in each cluster to estimate parameters related to the anomaly occurrence history. Figure 6 In the diagram, the eighth column indicates whether an anomaly occurred in each dataset. Controller 120 can estimate parameters associated with the anomaly occurrence history by comprehensively analyzing the anomaly history from the datasets and the classification to which the datasets belong. For example, among datasets F to J that share the same pattern 10112, controller 120 can infer that the anomalies occurring in datasets F, H, and J have the same or similar causes.

[0094] Although Figure 6 The anomaly occurrence history is simply represented as the occurrence of an anomaly, but the data management device 100 can include various information related to the anomaly in the anomaly occurrence history, such as the number of occurrences, the anomaly type, and the time of occurrence.

[0095] Controller 120 can also manage and store the location of each dataset or BMS that has been acquired, such as Figure 6 As shown in the seventh column.

[0096] Figure 7 This is a flowchart illustrating an operation method of a data analysis device according to an embodiment disclosed in this document. Figure 7 The embodiments shown are merely illustrative, and the operational sequence of various embodiments according to the present invention may vary. Figure 7 The order of operations shown is different, and Figure 7 Some of the operations shown can be omitted, rearranged, or combined.

[0097] Reference Figure 7The operation method of the data analysis device may include the following operations: in operation S100, acquiring multiple datasets, each dataset containing time series data of multiple parameters; in operation S200, extracting features of each dataset based on the correlation between the multiple parameters; in operation S300, extracting patterns of each dataset based on the similarity between the extracted features; in operation S400, classifying the multiple datasets into multiple clusters based on the patterns; and in operation S500, estimating parameters associated with the anomaly occurrence history by analyzing the anomaly occurrence history associated with the datasets included in each cluster.

[0098] In operation S100, the information acquisition unit 110 can acquire multiple datasets. The information acquisition unit 110 can acquire time-series data as a dataset from each of the multiple battery management systems.

[0099] In operation S200, controller 120 can extract features for each dataset based on the correlation between multiple parameters. For example, controller 120 can represent the features as a matrix indicating Pearson correlation coefficients of multiple parameters.

[0100] In operation S300, controller 120 can extract patterns for each dataset based on the similarity between the extracted features. For example, controller 120 can generate a correlation coefficient matrix from the features and apply a classification algorithm to the correlation coefficient matrix to extract patterns for each dataset.

[0101] In operation of S400, controller 120 can classify multiple datasets based on patterns. For example, controller 120 can classify datasets with the same pattern into the same cluster.

[0102] In operation S500, controller 120 can analyze the anomaly occurrence history associated with the datasets included in each cluster to estimate parameters associated with the anomaly occurrence history.

[0103] Figure 8 This is a flowchart illustrating the process of extracting a mode according to the implementation method disclosed in this document.

[0104] Reference Figure 8 The controller 120 can calculate the correlation coefficient matrix from multiple datasets and extract patterns from the correlation coefficient matrix.

[0105] In operation S310, controller 120 can calculate the correlation coefficients between features of multiple datasets to generate a correlation coefficient matrix composed of the correlation coefficients.

[0106] In operation S320, controller 120 can apply a classification algorithm to the correlation coefficient matrix to extract patterns. The classification algorithm may include, for example, the K-means algorithm. Controller 120 can apply the K-means algorithm while changing the K value on the correlation coefficient matrix and accumulate the classification results. Controller 120 can then extract patterns from each dataset from the accumulated classification results.

[0107] Figure 9 This is a block diagram illustrating the hardware configuration of a computing system for performing an operation method of a data analysis device according to an embodiment disclosed in this document.

[0108] Reference Figure 9 The computing system 1000 according to the embodiments disclosed in this document may include an MCU 1010, a memory 1020, an input / output interface (I / F) 1030, and a communication interface (I / F) 1040.

[0109] MCU 1010 can be a processor that executes various programs stored in memory 1020 (e.g., battery data acquisition program, feature extraction program, correlation coefficient matrix generation program, pattern extraction program, classification program, etc.), and processes various information, including time series data of the dataset, through these programs to perform reference... Figure 2 The data analysis equipment described includes the functions of the controller.

[0110] The memory 1020 can store various programs such as battery data acquisition programs, feature extraction programs, correlation coefficient matrix generation programs, pattern extraction programs, and classification programs. The memory 1020 can also store various information, including classification results of the dataset and anomaly occurrence history.

[0111] Multiple 1020s can be configured as needed. The memory 1020 can be volatile or non-volatile. As volatile memory, the memory 1020 can be RAM, DRAM, SRAM, etc. As non-volatile memory, the memory 1020 can be ROM, PROM, EAROM, EPROM, EEPROM, flash memory, etc. The examples of memory 1020 listed above are merely examples, and the memory is not limited to these examples.

[0112] The Input and Output I / F 1030 is an interface for connecting input devices (not shown) such as a keyboard, mouse, or touch panel, output devices such as a display (not shown), and the MCU 1010 to send and receive data.

[0113] The communication I / F 1040 can be a component capable of communicating various types of data with a server, including various devices that support wired or wireless communication. For example, a data analysis device can send various information, including the history of anomalies in the dataset, to a separate external server via the communication I / F 1040, and receive various information, including the history of anomalies in the dataset, from a separate external server.

[0114] Thus, the computer program according to the embodiments disclosed in this document can be recorded in the memory 1020 and processed by the MCU 1010 to serve as an execution reference. Figure 2 The module that describes the functionality.

[0115] Although all components are described as operating in combination or as a whole, the implementations disclosed in this document are not limited to these examples. That is, within the scope of the implementations disclosed in this document, all components may also be selectively combined in one or more ways to function.

[0116] Furthermore, unless otherwise expressly stated, terms such as “comprising,” “including,” or “having” used above should be interpreted as implying the inclusion of the specified components, rather than excluding other components, and may include additional components. Unless otherwise defined, all terms, including technical or scientific terms as used herein, have the same meaning as commonly understood by one of ordinary skill in the art to which the embodiments disclosed in this invention pertain. It will be further understood that terms such as those defined in commonly used dictionaries should be interpreted as having the meaning consistent with their meaning in the context of the relevant field, and unless expressly defined herein, these terms will not be interpreted in an idealized or overly formal sense.

[0117] The above description is merely an illustrative example of the technical ideas disclosed in this document, and those skilled in the art will be able to make various modifications and changes without departing from the subject matter of the embodiments disclosed in this document. Therefore, the embodiments disclosed in this document are not intended to limit the technical concept disclosed in this document, but are for illustrative purposes, and the scope of the technical concept disclosed in this document is not limited by these embodiments. The scope of protection of the technical concept disclosed in this document should be interpreted in accordance with the appended claims, and all equivalent technical concepts should be considered to be included within the scope of protection of this document.

[0118] [Explanation of reference numerals in the attached figures]

[0119] 10, 20, 30: Battery Management System

[0120] 100: Data analysis equipment

[0121] 110: Information Acquisition Unit

[0122] 120: Controller

[0123] 121: Feature Extraction Module

[0124] 123: Classification Module

[0125] 125: Cause Identification Module

[0126] 1000: Computing System

[0127] 1010: MCU

[0128] 1020: Memory

[0129] 1030: Input / Output I / F

[0130] 1040: Communication I / F

Claims

1. A data analysis device comprising: an information acquisition unit configured to acquire a plurality of data sets each containing time series data of a plurality of parameters; and a controller configured to extract a feature of each data set based on a correlation between the plurality of parameters, extract a pattern of each data set based on similarity between the extracted features, and classify the plurality of data sets into a plurality of clusters based on the pattern.

2. The data analysis device of claim 1, wherein, The controller extracts the feature as a matrix representing the correlation between the plurality of parameters.

3. The data analysis device of claim 2, wherein, The correlation between the plurality of parameters is represented by a Pearson correlation coefficient of each parameter pair.

4. The data analysis device of claim 1, wherein, The controller calculates a correlation coefficient between the features of the plurality of data sets and generates a correlation coefficient matrix composed of the correlation coefficients.

5. The data analysis device of claim 4, wherein, The controller applies a classification algorithm to the correlation coefficient matrix to extract the pattern.

6. The data analysis device of claim 5, wherein, The classification algorithm is a K-means algorithm.

7. The data analysis device of claim 6, wherein, The controller extracts the pattern based on a result of applying the classification algorithm to the correlation coefficient matrix while changing a value of K.

8. The data analysis device of claim 1, wherein, The controller analyzes a characteristic of each of the plurality of clusters.

9. The data analysis device of claim 8, wherein, The controller analyzes an abnormal occurrence history related to the data set included in each of the plurality of clusters to estimate the parameter associated with the abnormal occurrence history.

10. The data analysis device of claim 1, wherein, The controller extracts the feature by setting a window interval for each of the plurality of data sets. 11.An operation method of a data analysis device, the operation method of the data analysis device comprising the steps of: acquiring a plurality of data sets each containing time series data of a plurality of parameters; extracting a feature of each data set based on a correlation between the plurality of parameters; extracting a pattern of each data set based on similarity between the extracted features; and classifying the plurality of data sets into a plurality of clusters based on the pattern. The step of extracting the feature of each data set includes the step of extracting the feature as a matrix representing the correlation between the plurality of parameters.

12. The method of operating a data analysis device according to claim 11, wherein, The step of extracting the pattern of each data set includes the steps of:

13. The method of operating a data analysis device according to claim 11, wherein, calculating a correlation coefficient between the features of the plurality of data sets and generating a correlation coefficient matrix composed of the correlation coefficients; and applying a classification algorithm to the correlation coefficient matrix to extract the pattern. The classification algorithm is a K-means algorithm, and the step of extracting the pattern by applying the classification algorithm includes the step of extracting the pattern based on a result of applying the classification algorithm to the correlation coefficient matrix while changing a value of K.

14. The method of operating a data analysis device according to claim 13, wherein, The parameter associated with the abnormal occurrence history is estimated by analyzing an abnormal occurrence history related to the data set included in each of the plurality of clusters.

15. The method of operating a data analysis device according to claim 11, further comprising the step of: ​

Images