Battery management device and method

Abstract

A battery management device according to one embodiment of the present invention includes: a feature value extraction unit configured to extract feature values ​​of a learning battery for each charge / discharge cycle; a first state determination unit configured to determine the state of the learning battery for each charge / discharge cycle based on the feature values ​​extracted by the feature value extraction unit and a reference value previously set to correspond to the feature values; a labeling unit configured to set labels for the feature values ​​of the learning battery for each charge / discharge cycle based on the state of the learning battery determined by the first state determination unit; a model learning unit configured to train a classification model for determining the state of a battery for analysis based on the feature values ​​to which the labels are set by the labeling unit; and a second state determination unit configured to determine the state of the battery for analysis based on the classification model learned by the model learning unit.

Description

[Technical field]

[0001] This application claims priority to Korean Patent Application No. 10-2020-0161667, filed on November 26, 2020, the entire contents of which are incorporated herein by reference in their entirety in the specification and drawings.

[0002] The present invention relates to a battery management device and method, and more particularly to a battery management device and method capable of determining a battery state using a learned classification model. [Background technology]

[0003] In recent years, as the demand for portable electronic products such as laptops, video cameras, and mobile phones has grown rapidly and the development of electric vehicles, energy storage batteries, robots, satellites, and other products has progressed in earnest, active research is being conducted into high-performance batteries that can be repeatedly charged and discharged.

[0004] Currently, various types of batteries are commercially available, including nickel-cadmium batteries, nickel-metal hydride batteries, nickel-zinc batteries, and lithium-ion batteries. Among them, lithium-ion batteries are attracting attention due to their advantages over nickel-based batteries, such as almost no memory effect, easy charging and discharging, a very low self-discharge rate, and a high energy density.

[0005] Conventionally, to estimate or determine the state of such a battery, techniques based on machine learning or deep learning have been combined. For example, a technology has been developed to determine the state of a battery based on a trained model.

[0006] A large amount of learning data is required to train such a model, but in order to obtain the learning data from the battery, the battery must be charged and discharged, and it takes a long time to obtain the large amount of learning data. Furthermore, as the expected life of the battery increases, the time required for the charge and discharge cycle increases proportionally, so the time required to obtain the learning data also increases.

[0007] In addition, there is a limit to the amount of training data that can be obtained from a limited number of batteries. Therefore, there is a need to develop technology that can improve the performance of trained models by obtaining a large amount of training data even if the number of batteries is limited. Summary of the Invention [Problem to be solved by the invention]

[0008] The present invention has been made in consideration of the above problems, and aims to provide a battery management device and method that can secure a large amount of learning data for training a classification model from a limited number of batteries.

[0009] Other objects and advantages of the present invention can be understood from the following description and will become more apparent from the embodiments of the present invention. Also, the objects and advantages of the present invention can be realized by the means and combinations thereof as set forth in the claims. [Means for solving the problem]

[0010] A battery management device according to one embodiment of the present invention includes a feature value extraction unit configured to extract a feature value of a learning battery for each charge and discharge cycle, a first state judgment unit configured to judge a state of the learning battery for each charge and discharge cycle based on the feature value extracted by the feature value extraction unit and a reference value previously set to correspond to the feature value, a labeling unit configured to set a label for the feature value of the learning battery for each charge and discharge cycle based on the state of the learning battery judged by the first state judgment unit, a model learning unit configured to train a classification model for determining a state of a battery for analysis based on the feature value to which the label is set by the labeling unit, and a second state judgment unit configured to judge the state of the battery for analysis based on the classification model learned by the model learning unit.

[0011] The first state determination unit may be configured to determine the state of the learning battery as a normal state or a defective state for each charge / discharge cycle.

[0012] The labeling unit may be configured to set a first label to a characteristic value of the learning battery for a corresponding charge / discharge cycle when the learning battery is determined to be in a normal state.

[0013] The labeling unit may be configured to set a second label on a characteristic value of the learning battery for a corresponding charge / discharge cycle when the state of the learning battery is determined to be a defective state.

[0014] The labeling unit may be configured to set a label for a characteristic value of the learning battery based on a state of the learning battery determined in a predetermined charge / discharge cycle.

[0015] The labeling unit may be configured to set the first label to a characteristic value of the learning battery for charge / discharge cycles up to and including the specified charge / discharge cycle when the state of the learning battery is determined to be the normal state in the specified charge / discharge cycle.

[0016] The labeling unit may be configured to set the second label to a characteristic value of the learning battery for charge / discharge cycles after the predetermined charge / discharge cycle when the state of the learning battery is determined to be the defective state in the predetermined charge / discharge cycle.

[0017] The labeling unit may be configured to set the second label to a characteristic value of the learning battery for a subsequent charge / discharge cycle when the state of the learning battery is determined to be the defective state.

[0018] The labeling unit may be configured to count the number of times the state of the learning battery is determined to be the defective state, and if the counted number is equal to or greater than a reference number, to set the second label to the characteristic value of the learning battery for charge / discharge cycles subsequent to the last charge / discharge cycle in which the state of the learning battery was determined to be the defective state.

[0019] The feature extractor may be configured to extract a plurality of feature values ​​for the training battery.

[0020] The first state judgment unit may be configured to judge the state of the learning battery for each of the plurality of characteristic values ​​for each charge / discharge cycle based on a plurality of characteristic values ​​of the learning battery and a reference value preset to correspond to each of the plurality of characteristic values.

[0021] The labeling unit may be configured to set the label for each of the plurality of characteristic values ​​for each charge / discharge cycle.

[0022] The second state judgment unit may be configured to extract feature values ​​of the battery for analysis for each charge / discharge cycle and judge the state of the battery for analysis by inputting the extracted feature values ​​into the learned classification model.

[0023] A battery pack according to another aspect of the present invention includes a battery management device according to an aspect of the present invention.

[0024] According to yet another aspect of the present invention, a battery management method includes a feature value extraction step of extracting a feature value of a learning battery for each charge / discharge cycle; a first state determination step of determining a state of the learning battery for each charge / discharge cycle based on the feature value extracted in the feature value extraction step and a reference value preset to correspond to the feature value; a labeling step of setting a label for the feature value of the learning battery for each charge / discharge cycle based on the state of the learning battery determined in the first state determination step; a model learning step of training a classification model for determining a state of a battery for analysis based on the feature value to which the label is set in the labeling step; and a second state determination step of determining the state of the battery for analysis based on the classification model learned in the model learning step. Effect of the Invention

[0025] According to one aspect of the present invention, a large amount of training data for training a classification model can be quickly obtained from a limited number of batteries.

[0026] The effects of the present invention are not limited to the effects described above, and other effects of the present invention not mentioned will be clearly understood by those skilled in the art from the description of the claims.

[0027] The following drawings attached to this specification, together with the detailed description of the invention described below, serve to further understand the technical ideas of the present invention, and the present invention should not be interpreted as being limited only to the matters described in the drawings. [Brief description of the drawings]

[0028] [Figure 1] 1 is a diagram illustrating a battery management device according to an embodiment of the present invention; [Diagram 2] FIG. 2 is a diagram showing a first example in which a label is set by a battery management device according to an embodiment of the present invention. [Diagram 3] FIG. 11 is a diagram showing a second example of how a label is set by a battery management device according to an embodiment of the present invention. [Figure 4] FIG. 13 is a diagram showing a third example in which a label is set by a battery management device according to an embodiment of the present invention. [Diagram 5] FIG. 2 is a schematic diagram illustrating an exemplary configuration of a battery pack according to another embodiment of the present invention. [Figure 6] 13 is a diagram illustrating a battery management method according to yet another embodiment of the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0029] The terms and words used in this specification and the claims should not be interpreted in a limited manner to their general and dictionary meanings, but should be interpreted in a meaning and concept that corresponds to the technical idea of ​​the present invention, based on the principle that the inventor himself can appropriately define the concept of the term in order to explain the invention in the best possible way.

[0030] Therefore, it should be understood that the embodiment described in this specification and the configurations shown in the drawings are merely the most preferred embodiment of the present invention and do not represent the entire technical idea of ​​the present invention, and that there may be various equivalents and modifications that can be substituted for them at the time of this application.

[0031] In addition, in the description of the present invention, if it is determined that a detailed description of related known configurations or functions may make the gist of the present invention unclear, the detailed description will be omitted.

[0032] Terms including ordinal numbers, such as "first" and "second," are used to distinguish one of various components from other components, and are not intended to limit the components.

[0033] Throughout the specification, when a part "comprises" a certain element, this does not mean to exclude other elements, but means that it may further include other elements, unless otherwise specified.

[0034] Furthermore, throughout this specification, when a part is referred to as being "connected" to another part, this includes not only a "direct connection" but also an "indirect connection" through other elements.

[0035] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings.

[0036] FIG. 1 is a schematic diagram of a battery management system 100 according to an embodiment of the present invention.

[0037] Referring to FIG. 1, a battery management device 100 includes a feature value extracting unit 110, a first state determining unit 120, a labeling unit 130, a model learning unit 140, and a second state determining unit 150.

[0038] The feature value extractor 110 may be configured to extract the feature value of the learning battery for each charge / discharge cycle.

[0039] Here, the term "battery" refers to a single independent cell that has a negative terminal and a positive terminal and can be physically separated. For example, a pouch-type lithium ion cell may be a battery.

[0040] In addition, the training battery is a battery for obtaining a training data set for training a classification model, and can be distinguished from the analysis battery.

[0041] For example, the feature value extraction unit 110 may extract one or more feature values ​​for the learning battery for each charge / discharge cycle. Specifically, the feature value extraction unit 110 may extract feature values ​​for the learning battery, such as voltage, current, capacity, charging time, internal resistance, SOC (State of Charge), SOH (State of Health), or a combination thereof.

[0042] The first state determination unit 120 may be configured to determine the state of the learning battery for each charge / discharge cycle based on the feature values ​​extracted by the feature value extraction unit 110 and a reference value preset to correspond to the feature values.

[0043] Preferably, a reference value may be preset for each characteristic value in each charge / discharge cycle.

[0044] For example, assume that a first characteristic value to an Nth characteristic value for the learning battery are extracted in the first charge / discharge cycle to the Nth charge / discharge cycle, respectively. Here, N may be an integer equal to or greater than 2. In this case, a total of N reference values ​​may be set in advance. That is, the first reference value to the Nth reference value may be set in advance so as to correspond to the first characteristic value to the Nth characteristic value, respectively. Then, the first state determination unit 120 may compare the first characteristic value with the first reference value to determine the state of the learning battery in the first charge / discharge cycle. Similarly, the first state determination unit 120 may compare the Nth characteristic value with the Nth reference value to determine the state of the learning battery in the Nth charge / discharge cycle.

[0045] Specifically, the first state determination unit 120 may be configured to determine the state of the learning battery as a normal state or a defective state for each charge / discharge cycle.

[0046] For example, the reference value may be a threshold value for determining the state of the learning battery. The first state determination unit 120 may determine whether the learning battery is in a normal state or a defective state by comparing the reference value with the characteristic value.

[0047] The labeling unit 130 may be configured to set a label for the characteristic value of the learning battery for each charge / discharge cycle based on the state of the learning battery determined by the first state determination unit 120.

[0048] Specifically, the labeling unit 130 may be configured to set a first label to a characteristic value of the learning battery for a corresponding charge / discharge cycle when the learning battery is determined to be in a normal state.

[0049] In contrast, the labeling unit 130 may be configured to set a second label to the characteristic value of the learning battery for the corresponding charge / discharge cycle when the learning battery is determined to be in a defective state.

[0050] That is, the labeling unit 130 may set a label for the characteristic value of the learning battery extracted in each charge / discharge cycle according to the state of the learning battery determined by the first state determination unit 120.

[0051] The model training unit 140 may be configured to train a classification model for determining the state of the battery for analysis based on the feature values ​​labeled by the labeling unit 130.

[0052] For example, the classification model may be logistic regression, linear discriminant analysis, a support vector machine, a random forest classifier, or an artificial neural network.

[0053] The classification model trained by the model trainer 140 may be used to determine the state of the battery for analysis.

[0054] However, since the process of training a classification model based on labeled feature values ​​is a common process of model training in machine learning, a detailed description thereof will be omitted.

[0055] The second state determination unit 150 may be configured to determine the state of the analysis battery based on the classification model trained by the model training unit 140.

[0056] Specifically, the second state determination unit 150 may be configured to extract a characteristic value of the battery for analysis for each charge / discharge cycle.

[0057] Here, the characteristic value of the analysis battery may be the same as the characteristic value of the learning battery extracted by the characteristic value extraction unit 110. That is, the characteristic value of the analysis battery extracted by the second state determination unit 150 may be the same as the characteristic value used in the process in which the model learning unit 140 learns the classification model.

[0058] For example, assume that the feature value extraction unit 110 extracts the voltage of the learning battery for each charge / discharge cycle as a feature value. The first state determination unit 120 may determine the state of the learning battery for each charge / discharge cycle based on the voltage of the learning battery and a reference value. The labeling unit 130 may set a label for the voltage of each charge / discharge cycle based on the state of the learning battery determined for each charge / discharge cycle. The model learning unit 140 may train a classification model based on the voltages to which the labels are set. The second state determination unit 150 may extract the voltage of the analysis battery for each charge / discharge cycle as a feature value.

[0059] The second state determination unit 150 may be configured to input the extracted feature values ​​into the trained classification model to determine the state of the battery for analysis.

[0060] For example, the second state determination unit 150 may set the feature value of the battery for analysis as an input for the trained classification model. Then, the trained classification model may output the state of the battery for analysis as an output for the input. The second state determination unit 150 may obtain the output output from the trained classification model and determine the state of the battery for analysis as a normal state or a defective state.

[0061] The battery management device 100 according to an embodiment of the present invention has an advantage in that a large amount of learning data for training a classification model from a limited number of batteries can be secured.

[0062] That is, the battery degrades as the charge / discharge cycle progresses, and the battery state may differ in each charge / discharge cycle. The battery management device 100 extracts the feature value of the battery for each charge / discharge cycle in consideration of such battery characteristics, and can secure a large amount of learning data for training the classification model.

[0063] Therefore, even if the number of training batteries is limited, a large amount of training data for training the classification model can be secured, and the performance of the trained classification model can be improved based on such a large amount of training data, thereby improving the accuracy of the state determination of the analysis battery based on the trained classification model.

[0064] Meanwhile, referring to FIG. 1, the battery management device 100 may further include a recording unit 160. The recording unit 160 may store data required for each component of the battery management device 100 to operate and function, or data generated in the process of performing a program or an operation and function. The recording unit 160 is not particularly limited in type as long as it is a known information recording means known to be capable of recording, erasing, updating, and reading data. As an example, the information recording means may include a RAM, a flash memory, a ROM, an EEPROM, a register, etc. Also, the recording unit 160 may store a program code that defines a process executable by the control unit.

[0065] For example, the characteristic values ​​of the battery extracted for each charge / discharge cycle by the characteristic value extracting unit 110 may be stored in the recording unit 160. In addition, the label set for each characteristic value by the labeling unit 130 may also be stored in the recording unit 160.

[0066] Additionally, the feature value extractor 110 may be configured to extract a number of feature values ​​for the training battery.

[0067] Specifically, the feature value extractor 110 may extract two or more feature values ​​for each of the learning batteries for each charge / discharge cycle.

[0068] For example, the feature value extraction unit 110 may extract the voltage and internal resistance for each learning battery for each charge / discharge cycle. If 500 charge / discharge cycles are performed on 10 learning batteries, the feature value extraction unit 110 may extract a total of 10,000 feature values. That is, 5,000 voltages and 5,000 internal resistances may be extracted.

[0069] The first state judgment unit 120 may be configured to judge the state of the learning battery for each of the multiple characteristic values ​​for each charge / discharge cycle based on multiple characteristic values ​​of the learning battery and reference values ​​preset to correspond to each of the multiple characteristic values.

[0070] For example, in the above-described embodiment, if the feature value extraction unit 110 extracts a total of 10,000 feature values, the first state determination unit 120 can determine the state of the learning battery for the voltage and internal resistance for each charge / discharge cycle by comparing each of the 10,000 feature values ​​with the corresponding reference value. That is, the state based on the voltage of the learning battery in each of the 1st to 500th cycles can be determined based on the voltage of the learning battery in each of the 1st to 500th cycles by the voltage of the 10 learning batteries. Also, the state based on the internal resistance of the learning battery in each of the 1st to 500th cycles can be determined based on the internal resistance of the learning battery in each of the 1st to 500th cycles by the internal resistance of the 10 learning batteries.

[0071] The labeling unit 130 can be configured to set a label for each of the multiple characteristic values ​​for each charge / discharge cycle.

[0072] For example, in the above-described embodiment, when the first state determination unit 120 has determined the states of the learning battery for all of the 10,000 feature values, the labeling unit 130 may set a label for each of the 10,000 feature values.

[0073] That is, the battery management device 100 according to an embodiment of the present invention can obtain a large amount of learning data from a limited number of learning batteries by extracting a plurality of feature values ​​for the learning battery.

[0074] Various embodiments in which the labeling unit 130 sets labels for feature values ​​will now be described.

[0075] FIG. 2 is a diagram showing a first example of how a label is set by the battery management unit 100 according to an embodiment of the present invention.

[0076] The labeling unit 130 may be configured to set a label for the characteristic value of the learning battery based on the state of the learning battery determined in a predetermined charge / discharge cycle.

[0077] For example, in the embodiment of FIG. 2, the labeling unit 130 may set labels for the characteristic values ​​of each of the first learning battery B1 to the fifth learning battery B5 based on the state of each of the first learning battery B1 to the fifth learning battery B5 determined in the target cycle TC.

[0078] Here, the target cycle TC may be the last charge / discharge cycle at which the state of the battery is determined by the first state determination unit 120. For example, the target cycle TC may be a cycle related to quality assurance for the battery. As a specific example, if the quality assurance for the battery is set such that the battery capacity is preserved at 90% or more after 500 cycles, the target cycle TC may be set to 500 cycles. That is, the target cycle TC may be set in various ways depending on the type of battery, the location of use, the usage environment, etc.

[0079] Specifically, if the state of the learning battery is determined to be normal in a specified charge / discharge cycle, the labeling unit 130 may set a first label to the characteristic value of the learning battery for the charge / discharge cycle up to and including the specified charge / discharge cycle.

[0080] In contrast, if the state of the learning battery is determined to be defective in a certain charge / discharge cycle, the labeling unit 130 may set a second label to the characteristic value of the learning battery for charge / discharge cycles after the certain charge / discharge cycle.

[0081] 2, it is assumed that the states of the first learning battery B1 to the third learning battery B3 are determined to be normal and the states of the fourth learning battery B4 and the fifth learning battery B5 are determined to be defective at the target cycle TC. In this case, the states of the first learning battery B1 to the third learning battery B3 up until just before the target cycle TC should be normal, and the states of the fourth learning battery B4 and the fifth learning battery B5 after the target cycle TC should be defective.

[0082] The labeling unit 130 can set a first label to the characteristic values ​​of the first learning battery B1 to the third learning battery B3 up to just before the target cycle TC.

[0083] In addition, the labeling unit 130 may set the second label for the characteristic values ​​of the fourth learning battery B4 and the fifth learning battery B5 after the target cycle TC. That is, the labeling unit 130 may quickly set the label for the characteristic values ​​even if the states of the fourth learning battery B4 and the fifth learning battery B5 are not determined in the charge / discharge cycle after the target cycle TC.

[0084] That is, the battery management unit 100 can set labels for the characteristic values ​​of the learning battery not only up to the target cycle TC, but also for the charge / discharge cycles after the target cycle TC.

[0085] FIG. 3 is a diagram showing a second example of how a label is set by the battery management unit 100 according to an embodiment of the present invention.

[0086] The labeling unit 130 may be configured to set a second label on the characteristic value of the learning battery for subsequent charge / discharge cycles when the learning battery's condition is determined to be a defective state.

[0087] Specifically, the labeling unit 130 may set a second label to the characteristic value of the learning battery after the charge / discharge cycle in which the learning battery is determined to be in a defective state. That is, even if the first state determination unit 120 does not determine the state of the learning battery after the charge / discharge cycle in which the learning battery is determined to be in a defective state, the labeling unit 130 may set a label to the characteristic value.

[0088] For example, in the embodiment of Fig. 3, suppose that the state of the sixth learning battery B6 at the target cycle TC is determined to be normal, and the state of the seventh learning battery B7 is determined to be defective. Suppose that the state of the eighth learning battery B8 at the third charge / discharge cycle C3 is determined to be defective, and the state of the ninth learning battery B9 at the second charge / discharge cycle C2 is determined to be defective. Suppose that the state of the tenth learning battery B10 at the first charge / discharge cycle C1 is determined to be defective.

[0089] The labeling unit 130 may set a first label for the feature values ​​of the sixth learning battery B6 up to just before the target cycle TC.

[0090] Furthermore, the labeling unit 130 may set a first label to the feature values ​​of the seventh learning battery B7 up until just before the target cycle TC, and may set a second label to the feature values ​​after the target cycle TC. That is, even if the state of the seventh learning battery B7 is not determined by the first state determination unit 120 from the target cycle TC, the labeling unit 130 may set the second label to the feature values ​​of the seventh learning battery B7.

[0091] Furthermore, the labeling unit 130 may set a first label to the characteristic values ​​of the eighth learning battery B8 up until just before the third charge / discharge cycle C3, and may set a second label to the characteristic values ​​after the third charge / discharge cycle C3. That is, even if the state of the eighth learning battery B8 is not determined by the first state determination unit 120 from the third charge / discharge cycle C3 onwards, the labeling unit 130 may set the second label to the characteristic values ​​of the eighth learning battery B8.

[0092] Furthermore, the labeling unit 130 may set a first label to the characteristic values ​​of the ninth learning battery B9 up until just before the second charging / discharging cycle C2, and may set a second label to the characteristic values ​​after the second charging / discharging cycle C2. That is, even if the state of the ninth learning battery B9 is not determined by the first state determination unit 120 from the second charging / discharging cycle C2 onwards, the labeling unit 130 may set the second label to the characteristic values ​​of the ninth learning battery B9.

[0093] Furthermore, the labeling unit 130 may set a first label to the characteristic values ​​of the tenth learning battery B10 up to just before the first charge / discharge cycle C1, and may set a second label to the characteristic values ​​after the first charge / discharge cycle C1. That is, even if the state of the tenth learning battery B10 is not determined by the first state determination unit 120 from the first charge / discharge cycle C1 onwards, the labeling unit 130 may set a second label to the characteristic values ​​of the tenth learning battery B10.

[0094] The battery management device 100 according to an embodiment of the present invention may not determine the state of the learning battery after the charge / discharge cycle in which the learning battery is determined to be in a defective state, and may set a second label to the feature value. Therefore, learning data for training the classification model may be obtained more quickly, and system resources of the battery management device 100 may be saved in the process of obtaining the learning data.

[0095] FIG. 4 is a diagram showing a third example in which a label is set by the battery management unit 100 according to an embodiment of the present invention.

[0096] The labeling unit 130 may be configured to count the number of times the state of the learning battery is determined to be a defective state.

[0097] For example, in the embodiment of Fig. 4, assume that the eleventh learning battery B11 is determined to be in a defective state in the first charge / discharge cycle C1. The labeling unit 130 may set the number of times that the eleventh learning battery B11 is determined to be in a defective state in the first charge / discharge cycle C1 to 1.

[0098] In addition, the eleventh learning battery B11 may be determined to be in a defective state in the second charge / discharge cycle C2. The labeling unit 130 may set the number of times that the eleventh learning battery B11 has been determined to be in a defective state in the second charge / discharge cycle C2 to 2.

[0099] Furthermore, the eleventh learning battery B11 may be determined to be in a defective state in the third charge / discharge cycle C3. The labeling unit 130 may set the number of times the eleventh learning battery B11 has been determined to be in a defective state in the third charge / discharge cycle C3 to 3.

[0100] The labeling unit 130 may be configured to set a second label to the characteristic value of the learning battery for charge / discharge cycles subsequent to the last charge / discharge cycle in which the state of the learning battery was determined to be defective if the counted number is equal to or greater than a reference number.

[0101] For example, in the embodiment of FIG. 4, it is assumed that the reference number is preset to 3. Since the number of times that the eleventh learning battery B11 is determined to be in a defective state in the third charge / discharge cycle C3 is set to 3, the labeling unit 130 may set the second label to the feature values ​​after the third charge / discharge cycle C3.

[0102] Generally, the battery condition deteriorates as the charge / discharge cycle is performed, but in some cases, the battery condition may temporarily recover. In addition, the process of determining the battery condition may be affected by temporary noise due to measurement errors caused by internal and external environmental issues. Therefore, the labeling unit 130 may set the second label even if the learning battery condition is not determined, only when the learning battery condition is determined to be defective for a reference number of times or more.

[0103] Meanwhile, a predetermined condition may be further added to the reference number. For example, if the state of the learning battery is continuously determined to be a defective state and the number of times the learning battery is determined to be a defective state is equal to or greater than the reference number, the labeling unit 130 may be configured to set a second label to the characteristic value of the learning battery for charge / discharge cycles subsequent to the last charge / discharge cycle in which the learning battery was determined to be a defective state.

[0104] The reference number and conditions that can be added to the reference number can be preset by a user in consideration of the type of battery, the location of use, the environment in which the battery is used, and the like.

[0105] Meanwhile, the feature value extractor 110 may be configured to generate additional feature values ​​based on the feature values ​​extracted for each charge / discharge cycle.

[0106] For example, assume that a total of 10 learning batteries are provided, and the feature value extracting unit 110 extracts feature values ​​for each learning battery from the first charge / discharge cycle C1 to just before the 500th charge / discharge cycle. In this case, a total of 5,000 feature values ​​can be extracted.

[0107] The feature value extraction unit 110 may generate additional feature values ​​by combining the feature values ​​extracted for each charge / discharge cycle to obtain more training data for training the classification model.

[0108] For example, the feature value extractor 110 may generate a plurality of groups for a plurality of feature values ​​extracted in the same charge / discharge cycle, and calculate an average feature value for each of the generated groups.

[0109] Additionally, the feature value extraction unit 110 may be configured to generate groups of like feature values.

[0110] For example, assume that voltage and internal resistance are extracted as feature values ​​by the feature value extraction unit 110. In this case, a voltage group may be generated based on the feature value for voltage, and an internal resistance group may be generated based on the feature value for internal resistance. That is, a group in which voltage and internal resistance are mixed is not generated.

[0111] In the above embodiment, the feature value extractor 110 may generate groups of 2 to 10 of the 10 feature values ​​extracted in the same charge / discharge cycle, and may further generate an average feature value for each group. Here, the number of groups that can be generated from the 10 feature values ​​may be 1,013 in total.

[0112] Specifically, when the feature value extraction unit 110 generates groups by combining 2, 3, 4, 5, 6, 7, 8, 9, and 10 of the 10 feature values, the numbers of groups generated may be 45, 120, 210, 252, 210, 120, 45, 10, and 1, respectively. Therefore, a total of 1,013 groups may be generated. The feature value extraction unit 110 may generate 1,013 additional feature values ​​for each charge / discharge cycle by calculating an average feature value for each of the 1,013 generated groups.

[0113] The first state determination unit 120 may determine a state of the virtual learning battery based on each of the additional feature values ​​generated by the feature value extraction unit 110. Then, the labeling unit 130 may set a label for each of the additional feature values ​​based on the determined state of the virtual learning battery. The model learning unit 140 may train a classification model by using all of the feature values ​​for the learning battery and the feature values ​​for the virtual learning battery.

[0114] That is, the battery management device 100 can secure a larger amount of learning data based on the limited number of learning batteries by generating additional feature values ​​for the virtual learning batteries.

[0115] Meanwhile, in the process of generating the groups, the feature value extraction unit 110 may be configured to further consider the state of the learning battery determined by the first state determination unit 120. That is, the feature value extraction unit 110 may be configured to generate a group by combining learning batteries having the same state determined by the first state determination unit 120.

[0116] For example, assume that in the N-th charge / discharge cycle, the states of six of the ten learning batteries are determined to be normal and the states of the four learning batteries are determined to be defective. The feature value extraction unit 110 may generate a plurality of groups by combining the feature values ​​of the six learning batteries determined to be normal. In this case, a total of 57 groups may be generated. The feature value extraction unit 110 may also generate a plurality of groups by combining the feature values ​​of the four learning batteries determined to be defective. In this case, a total of 11 groups may be generated. Then, the feature value extraction unit 110 may generate a total of 68 additional feature values ​​by calculating an average feature value for each of the generated groups.

[0117] The battery management device 100 according to the present invention is applicable to a BMS (Battery Management System). That is, the BMS according to the present invention may include the above-mentioned battery management device 100. In this configuration, at least a part of each component of the battery management device 100 may be implemented by complementing or adding functions of components included in a conventional BMS. For example, the feature value extraction unit 110, the first state determination unit 120, the labeling unit 130, the model learning unit 140, and the second state determination unit 150 of the battery management device 100 may be implemented as components of a BMS.

[0118] FIG. 5 is a diagram illustrating an exemplary configuration of a battery pack 1 according to another embodiment of the present invention.

[0119] 5, the battery management device 100 according to the present invention may be provided in a battery pack 1. That is, the battery pack 1 according to the present invention may include the above-mentioned battery management device 100 and one or more battery modules 10. Here, the battery module 10 may include one or more battery cells connected in series and / or parallel. In addition, the battery pack 1 may further include electrical components (relays, fuses, etc.), a case, etc.

[0120] 5, the feature value extraction unit 110 may be connected to the battery module 10 and the current measurement unit. The feature value extraction unit 110 may measure the voltage, current, temperature, etc. of each battery cell included in the battery module 10 and extract them as feature values. In addition, the feature value extraction unit 110 may extract the internal resistance, SOC, SOH, etc. of each battery cell based on the measured values ​​as feature values.

[0121] FIG. 6 is a diagram illustrating a battery management method according to yet another embodiment of the present invention.

[0122] Preferably, each step of the battery management method can be performed by the battery management device 100. In the following, the contents overlapping with the above description will be omitted or will be briefly described.

[0123] The feature value extraction step S100 is a step of extracting feature values ​​of the learning battery for each charge / discharge cycle, and can be performed by the feature value extraction unit 110.

[0124] The first state determination step S200 is a step of determining the state of the learning battery for each charge / discharge cycle based on the feature values ​​extracted in the feature value extraction step S100 and reference values ​​preset to correspond to the feature values, and can be performed by the first state determination unit 120.

[0125] For example, assume that a first characteristic value to an Nth characteristic value for the learning battery are extracted in the first charge / discharge cycle to the Nth charge / discharge cycle, respectively. Here, N may be an integer equal to or greater than 2. In this case, a total of N reference values ​​may be set in advance. That is, a first reference value to an Nth reference value may be set in advance so as to correspond to the first characteristic value to the Nth characteristic value, respectively. Then, the first state determination unit 120 may compare the first characteristic value with the first reference value to determine the state of the learning battery in the first charge / discharge cycle. Similarly, the first state determination unit 120 may compare the Nth characteristic value with the Nth reference value to determine the state of the learning battery in the Nth charge / discharge cycle.

[0126] The labeling step S300 is a step of setting a label for the characteristic value of the learning battery for each charge / discharge cycle based on the state of the learning battery determined in the first state determination step S200, and can be performed by the labeling unit 130.

[0127] Specifically, the labeling unit 130 may be configured to set a first label to a characteristic value of the learning battery for a corresponding charge / discharge cycle when the learning battery is determined to be in a normal state.

[0128] In contrast, the labeling unit 130 may be configured to set a second label to the characteristic value of the learning battery for the corresponding charge / discharge cycle when the learning battery is determined to be in a defective state.

[0129] The model learning step S400 is a step of training a classification model for determining the state of a battery to be analyzed based on the feature values ​​labeled in the labeling step S300, and can be performed by the model learning unit 140.

[0130] The second state determination step S500 is a step of determining the state of the analysis battery based on the classification model learned in the model learning step S400, and can be performed by the second state determination unit 150.

[0131] That is, the battery management method according to an embodiment of the present invention has an advantage in that a large amount of learning data for learning a classification model from a limited number of batteries can be secured. Therefore, the performance of the trained classification model can be improved. Therefore, the accuracy of determining the state of a battery for analysis based on the trained classification model according to an embodiment of the present invention can be improved.

[0132] The above-described embodiments of the present invention may be embodied not only by an apparatus and a method, but also through a program that realizes functions corresponding to the configuration of the embodiments of the present invention or a recording medium on which the program is recorded, and such implementation would be easily embodied by a person skilled in the art from the description of the above-described embodiments.

[0133] As described above, the present invention has been described using limited embodiments and drawings. However, the present invention is not limited thereto, and it goes without saying that various modifications and variations can be made by a person having ordinary knowledge in the technical field to which the present invention belongs within the technical spirit of the present invention and the equivalent scope of the claims.

[0134] Furthermore, the present invention described above is susceptible to various substitutions, modifications and alterations by those having ordinary knowledge in the technical field to which the present invention pertains, within the scope not departing from the technical spirit of the present invention, and is not limited to the above-described embodiments and the accompanying drawings, but may be configured by selectively combining all or part of each embodiment for various modifications. [Explanation of symbols]

[0135] 1: Battery pack 10: Battery module 100: Battery management device 110: Feature value extraction unit 120: First state determination unit 130: Labeling Department 140: Model learning section 150: Second state determination unit 160: Recording section

Claims

1. A battery management device including a feature value extraction unit, a first state determination unit, a labeling unit, a model learning unit, and a second state determination unit, When a learning battery is connected to the battery management device, the feature value extraction unit is configured to extract a feature value of the learning battery for each charge / discharge cycle; the first state determination unit is configured to determine a state of the learning battery for each charge / discharge cycle based on the feature value extracted by the feature value extraction unit and a reference value preset to correspond to the feature value, the labeling unit is configured to set a label for a characteristic value of the learning battery for each charge / discharge cycle based on the state of the learning battery determined by the first state determination unit, The model learning unit is configured to train a classification model for determining a state of a battery for analysis based on the feature value to which the label is set by the labeling unit; When the analysis battery is connected to the battery management device, the second state determination unit is configured to determine a state of the analysis battery based on the classification model learned by the model learning unit; The labeling unit includes: If the state of the learning battery is determined to be normal, a first label is set to a characteristic value of the learning battery for a corresponding charge / discharge cycle; If the state of the learning battery is determined to be a defective state, a second label is set to the characteristic value of the learning battery for the corresponding charge / discharge cycle; and setting the second label to a characteristic value of the learning battery for subsequent charge / discharge cycles when the learning battery is determined to be in the defective state. Battery management device.

2. A battery management device including a feature value extraction unit, a first state determination unit, a labeling unit, a model learning unit, and a second state determination unit, When a learning battery is connected to the battery management device, the feature value extraction unit is configured to extract a feature value of the learning battery for each charge / discharge cycle; the first state determination unit is configured to determine a state of the learning battery for each charge / discharge cycle based on the feature value extracted by the feature value extraction unit and a reference value preset to correspond to the feature value, the labeling unit is configured to set a label for a characteristic value of the learning battery for each charge / discharge cycle based on the state of the learning battery determined by the first state determination unit, The model learning unit is configured to train a classification model for determining a state of a battery for analysis based on the feature value to which the label is set by the labeling unit; When the analysis battery is connected to the battery management device, the second state determination unit is configured to determine a state of the analysis battery based on the classification model learned by the model learning unit; The labeling unit includes: If the state of the learning battery is determined to be normal, a first label is set to a characteristic value of the learning battery for a corresponding charge / discharge cycle; If the state of the learning battery is determined to be a defective state, a second label is set to the characteristic value of the learning battery for the corresponding charge / discharge cycle; The labeling unit includes: a battery management device configured to count the number of times the state of the learning battery is judged to be in the defective state, and if the counted number is equal to or greater than a reference number, to set the second label to the characteristic value of the learning battery for charge / discharge cycles after the last charge / discharge cycle in which the state of the learning battery was judged to be in the defective state.

3. The labeling unit includes:

3. A battery management device as described in claim 1 or 2, configured to set the first label to the characteristic value of the learning battery for charge / discharge cycles up to just before the target charge / discharge cycle when the state of the learning battery is determined to be normal in a target charge / discharge cycle, which is the last charge / discharge cycle among a plurality of charge cycles in which the state of the learning battery is determined.

4. The feature value extraction unit configured to extract a plurality of feature values ​​for the learning battery when the learning battery is connected to the battery management device; The first state determination unit is 3. The battery management device of claim 1, further comprising: a battery management unit configured to determine a state of the learning battery for each of the plurality of characteristic values ​​for each charge / discharge cycle based on a plurality of characteristic values ​​of the learning battery and a reference value preset to correspond to each of the plurality of characteristic values.

5. The labeling unit includes: The battery management device according to claim 4 , configured to set the label for each of the plurality of characteristic values ​​for each of the charge / discharge cycles.

6. The second state determination unit is 3. The battery management device of claim 1, wherein when the analysis battery is connected to the battery management device, the device is configured to extract feature values ​​of the analysis battery for each charge / discharge cycle and input the extracted feature values ​​into the learned classification model to determine the state of the analysis battery.

7. A feature value extraction stage for extracting feature values ​​of a learning battery for each charge / discharge cycle; a first state determination step of determining a state of the learning battery for each charge / discharge cycle based on the feature value extracted in the feature value extraction step and a reference value preset to correspond to the feature value; a labeling step of setting a label for a characteristic value of the learning battery for each charge / discharge cycle based on the state of the learning battery determined in the first state determination step; a model learning step of learning a classification model for determining a state of a battery to be analyzed based on the feature values ​​to which the labels are set in the labeling step; and a second state determination step of determining a state of the analysis battery based on the classification model learned in the model learning step, The labeling step includes: If the state of the learning battery is determined to be normal, a first label is set to a characteristic value of the learning battery for a corresponding charge / discharge cycle; If the state of the learning battery is determined to be a defective state, a second label is set to the characteristic value of the learning battery for the corresponding charge / discharge cycle; If the state of the learning battery is determined to be the defective state, the second label is set to the characteristic value of the learning battery for subsequent charge / discharge cycles. Battery management methods.

8. A feature value extraction stage for extracting feature values ​​of a learning battery for each charge / discharge cycle; a first state determination step of determining a state of the learning battery for each charge / discharge cycle based on the feature value extracted in the feature value extraction step and a reference value preset to correspond to the feature value; a labeling step of setting a label for a characteristic value of the learning battery for each charge / discharge cycle based on the state of the learning battery determined in the first state determination step; a model learning step of learning a classification model for determining a state of a battery to be analyzed based on the feature values ​​to which the labels are set in the labeling step; and a second state determination step of determining a state of the analysis battery based on the classification model learned in the model learning step, The labeling step includes: If the state of the learning battery is determined to be normal, a first label is set to a characteristic value of the learning battery for a corresponding charge / discharge cycle; If the state of the learning battery is determined to be a defective state, a second label is set to the characteristic value of the learning battery for the corresponding charge / discharge cycle; the number of times that the state of the learning battery is determined to be the defective state is counted, and when the counted number is equal to or greater than a reference number, the second label is set to the characteristic value of the learning battery for charge / discharge cycles subsequent to the last charge / discharge cycle in which the state of the learning battery is determined to be the defective state. Battery management methods.

Images