Battery capacity measurement device and method, and battery control system including the same

Abstract

The present application relates to a battery capacity measuring device and method. The battery capacity measuring device includes a learning data input unit that receives battery capacity factor learning data measured in a certain time's charge and discharge process of a single battery selected as a learning target, a measurement data input unit that receives battery capacity factor measurement data selected in a certain time's charge and discharge process of a battery selected as a prediction target, a data learning unit that derives a battery capacity distribution from the battery capacity factor learning data input to the learning data input unit and performs a plurality of different machine learning for each battery capacity range of the battery capacity distribution derived from the learning data, and an output unit that calculates capacity prediction data of the battery selected as the prediction target based on the results of the plurality of different machine learning from the input battery capacity factor measurement data and outputs the battery capacity prediction data calculated for each battery capacity range of the battery capacity distribution derived from the learning data.

Description

TECHNICAL FIELD

[0001] This application claims priority to and the benefit of Korean Patent Application No. 10-2020-0151880, filed on November 13, 2020, in the Korean Intellectual Property Office, the entire contents of which are incorporated herein by reference.

[0002] The present application relates to an apparatus and a method for measuring a battery capacity.

[0003] The present application relates to a battery management system apparatus including an apparatus for measuring a battery capacity.

[0004] The present application relates to a mobile device including a battery management system apparatus.

[0005] The present application relates to a computer program stored in a recording medium for executing a method for measuring a battery capacity. BACKGROUND

[0006] The demand for secondary batteries is rapidly expanding for electric vehicles, mobile devices, etc., and the requirements for the state diagnosis and quality stability of secondary batteries are also increasing.

[0007] When the capacity of a secondary battery is determined to satisfy a rated capacity, which is a predetermined standard, by sampling a certain number of batteries from a battery tray generated from a cycle charging / discharging process and continuously charging and discharging the sampled batteries, a pass is made by the determination.

[0008] The capacity of such a secondary battery is expressed in a discrete form having a certain degree of deviation within the upper and lower limit ranges of the rated capacity, and depending on the manufacturing method and manufacturing conditions of the battery and the measurement conditions of the temperature, humidity, charging rate, and discharging rate to be measured, the results can deviate.

[0009] Although deviations can occur depending on these measurement conditions, the capacity value to be measured during charging or discharging is usually measured under conditions different from the operating conditions specified in the battery specification.

[0010] In addition, even if the battery is determined to be a good product as a result of the capacity measurement, various situations can occur when a module or a pack composed of a plurality of batteries corresponding to the good product is actually operated or used (charged and discharged).

[0011] That is, even when the module or the pack is actually operated or used, continuous battery capacity measurement and state diagnosis are required.

[0012] However, similar to in the production process, even during the capacity measurement and the state diagnosis of the battery used in the module or the pack, the results can deviate depending on the measurement conditions such as the temperature, humidity, charging rate, and discharging rate to be measured.

[0013] Therefore, in the production process of a battery and the process of actually operating a module or a group composed of a plurality of batteries, there is a need for a method of determining the quality of a battery and diagnosing the condition of a battery by monitoring the state of the battery in real time, correcting the deviation between capacity measurement values due to different measurement conditions, and then measuring the capacity of a secondary battery, and an apparatus capable of implementing the corresponding method.

[0014] <Patent Document> Korean Patent Application Publication No. 10-2004-0051195 SUMMARY

[0015] <Technical Problem>

[0016] An object of the present application is to provide an apparatus and a method capable of measuring the capacity of a battery by correcting the influence of the use conditions of the battery.

[0017] <Technical Solution>

[0018] One embodiment of the present disclosure provides an apparatus for measuring the capacity of a battery, the apparatus including: a learning data input unit for receiving battery capacity factor learning data in a specific time of a charge and discharge process of a battery selected as a learning target; a measurement data input unit for receiving selected battery capacity factor measurement data in a specific time of a charge and discharge process of a battery selected as a prediction target; a data learning unit for deriving a battery capacity distribution from the battery capacity factor learning data input to the learning data input unit and performing a plurality of different machine learnings for each battery capacity range of the battery capacity distribution derived from the learning data, respectively; and an output unit for calculating battery capacity prediction data from the input battery capacity factor measurement data and outputting the battery capacity prediction data calculated for each battery capacity range of the battery capacity distribution derived from the learning data by the results of the plurality of different machine learnings, respectively.

[0019] Another embodiment of the present disclosure provides a method for measuring the capacity of a battery, the method including the steps of: inputting battery capacity factor learning data in a specific time of a charge and discharge process of a battery selected as a learning target; deriving a battery capacity distribution from the input battery capacity factor learning data; performing a plurality of different machine learnings for each battery capacity range of the battery capacity distribution derived from the learning data, respectively; inputting selected battery capacity factor measurement data in a specific time of a charge and discharge process of a battery selected as a prediction target; calculating battery capacity prediction data from the input battery capacity factor measurement data; and outputting the battery capacity prediction data calculated for each battery capacity range of the battery capacity distribution derived from the learning data by the results of the plurality of machine learnings, respectively.

[0020] One embodiment of the present disclosure provides a battery management system device including the above-described device for measuring battery capacity.

[0021] One embodiment of the present disclosure relates to a mobile device including the above-described battery management system device.

[0022] Finally, one embodiment of the present disclosure relates to a computer program stored in a recording medium for executing the above-described method for measuring battery capacity.

[0023]

Advantageous Effects

[0024] The device and method for measuring battery capacity according to the embodiments of the present application can improve the accuracy of battery capacity measurement by correcting the effects of battery usage conditions, and can reduce the costs of battery manufacturing and capacity measurement in the process.

[0025] The device and method for measuring battery capacity according to the embodiments of the present application can improve the accuracy of diagnosing battery status and predicting battery life by improving the accuracy of battery capacity measurement.

[0026] The device and method for measuring battery capacity according to the embodiments of the present application can improve the efficiency of battery product quality control and the efficiency of battery activation processes by providing a cost-effective and accurate capacity measurement method and improving the accuracy of battery capacity measurement.

[0027] When a plurality of batteries in the form of modules, groups, and trays are mounted on an electric vehicle, a mobile device, etc., and used as a power source, the device and method for measuring battery capacity according to the embodiments of the present application can improve the accuracy of measuring the capacity of individual batteries when performing capacity matching and cell balancing, and can improve the life of batteries in the form of modules, groups, and trays. BRIEF DESCRIPTION OF DRAWINGS

[0028] Figure 1 FIG. 1 is a diagram showing a process of applying the device and method for measuring battery capacity according to the present application.

[0029] Figure 2 FIG. 2 is a diagram schematically showing the configuration of the device for measuring battery capacity according to the present application.

[0030] Figure 3 and Figure 4 are capacity distribution comparison diagrams and box plots showing the results of derivation according to the embodiments and comparative embodiments, respectively. DETAILED DESCRIPTION

[0031] Hereinafter, the present disclosure will be described in detail so that those of ordinary skill in the art can easily implement the present disclosure. However, the present disclosure can be embodied in various different forms, and is not limited to the configurations described herein.

[0032] In the present specification, if a prescribed part "includes" a prescribed element, this means that another element can be further included, rather than excluding other elements, unless there is any particularly opposite description.

[0033] In the present specification, the meaning of "at least one of" is that one or more and all of the following, for example, the meaning of "at least one of A, B, and C" is that it includes all cases of including one of A, B, or C, two of A and B, A and C, and B and C, and three (all) of A, B, and C.

[0034] That is, in the present specification, "learning data" means data for machine learning.

[0035] Further, in the present specification, "measurement data" means data to be inputted in order to calculate "prediction data", and the prediction data means data to be outputted as a result of machine learning performed on inputted measurement data.

[0036] Further, in the present specification, "capacity factor learning data" is learning data for obtaining accurate capacity measurement results using machine learning, and means data measured, collected, and stored in a state in which a battery is charged, discharged, and rested, including a battery charging voltage, a battery discharging voltage, a battery charging current, a battery discharging current, a battery charging capacity, a battery discharging capacity, a battery impedance, a battery temperature, etc., corresponding to a capacity measurement value of a single battery rated capacity. However, the capacity factor learning data is not limited to including the above factors, and can include all factors that affect the capacity of the battery and can be measured and collected.

[0037] Further, in the present specification, "capacity factor measurement data" is measurement data for obtaining accurate capacity measurement results using machine learning, and means data measured, collected, and stored in a state in which a battery is charged, discharged, and rested, including a battery charging voltage, a battery discharging voltage, a battery charging current, a battery discharging current, a battery charging capacity, a battery discharging capacity, a battery impedance, a battery temperature, etc. The capacity measurement data is not limited to the capacity factor measurement data including the above factors, and can include all factors that affect the capacity of the battery and can be measured and collected.

[0038] One embodiment of the present disclosure provides an apparatus for measuring battery capacity, the apparatus including: a learning data input unit for receiving battery capacity factor learning data in a certain time of a charge and discharge process of a battery selected as a learning target; a measurement data input unit for receiving battery capacity factor measurement data selected in a certain time of a charge and discharge process of a battery selected as a prediction target; a data learning unit for deriving a battery capacity distribution from the battery capacity factor learning data input to the learning data input unit, and performing a plurality of different machine learnings for each battery capacity range of the battery capacity distribution derived from the learning data, respectively; and an output unit for calculating battery capacity prediction data from the input battery capacity factor measurement data, and outputting the battery capacity prediction data calculated for each battery capacity range of the battery capacity distribution derived from the learning data by the results of the plurality of different machine learnings, respectively.

[0039] The battery capacity distribution is obtained as follows. After dividing the entire capacity range at constant or varying intervals between sections, the number of batteries having a corresponding capacity for each section is measured, and is displayed in the form of a bar graph.

[0040] In one embodiment of the present disclosure, the battery selected as the learning target and the battery selected as the prediction target can refer to individual batteries each independently disposed on a module, a group, and a tray.

[0041] Here, machine learning is a field of artificial intelligence, and refers to a technology in which a computer program improves information processing ability by learning using data and processing experience, or a technology related thereto. The technology related to machine learning is well known in the technical field to which the present disclosure pertains. That is, a detailed description of a specific learning algorithm of machine learning will be omitted.

[0042] In the present specification, "performing a plurality of different machine learnings for each battery capacity range of the battery capacity distribution derived from the learning data" refers to designating a battery capacity range as a certain standard in a battery capacity distribution, and performing different machine learnings for each range.

[0043] In one embodiment of the present disclosure, the battery can be a secondary battery, but the present disclosure is not limited thereto.

[0044] In the present specification, the term "certain time" refers to a time in which a charge and discharge process of an arbitrarily determined battery is performed. For example, when a charge and discharge process of a battery is to be performed for 1 hour, the certain time refers to 1 hour.

[0045] In one embodiment of the disclosure, the battery capacity factor learning data of the device for measuring the battery capacity can include the battery charge capacity and the battery discharge capacity measured during charging, discharging, and resting of the battery by the capacity measurement value corresponding to the rated capacity of the single battery selected as a learning target, and can further include one or more of the battery charge voltage, the battery discharge voltage, the battery open circuit voltage (OCV), the battery charge current, the battery discharge current, the battery impedance, and the battery temperature, but the disclosure is not limited thereto, and any factor that can affect the battery capacity can be included herein.

[0046] In one embodiment of the disclosure, the battery capacity factor learning data of the device for measuring the battery capacity can include the battery charge voltage, the battery discharge voltage, the battery open circuit voltage (OCV), the battery charge current, the battery discharge current, the battery charge capacity, the battery discharge capacity, the battery impedance, and the battery temperature measured during charging, discharging, and resting of the battery by the capacity measurement value corresponding to the rated capacity of the single battery selected as a learning target.

[0047] In one embodiment of the disclosure, the battery capacity factor measurement data of the device for measuring the battery capacity can include one or more of the battery charge voltage, the battery discharge voltage, the battery open circuit voltage (OCV), the battery charge current, the battery discharge current, the battery charge capacity, the battery discharge capacity, the battery impedance, and the battery temperature measured during charging, discharging, and resting of the battery selected as a prediction target.

[0048] In one embodiment of the disclosure, the battery capacity factor measurement data of the device for measuring the battery capacity can include the battery charge voltage, the battery discharge voltage, the battery open circuit voltage (OCV), the battery charge current, the battery discharge current, the battery charge capacity, the battery discharge capacity, the battery impedance, and the battery temperature measured during charging, discharging, and resting of the battery selected as a prediction target.

[0049] That is, the capacity measurement value of the rated capacity can correspond to a dependent variable, and an independent variable for estimating the dependent variable can be referred to as a capacity factor.

[0050] In this specification, different machine learning models refer to different machine learning algorithms, for example, a decision tree and a support vector machine (SVM). However, even if the same decision tree algorithm is used, when attributes (hyperparameters) representing the structure of the decision tree, such as the depth of the tree, the number of leaf nodes, and the like, are different from each other, or when the structure of the deep neural network, such as the input layer, the hidden layer, the output layer, the weight of each node, and the like, are different from each other, they are considered to be different machine learning models.

[0051] In one embodiment of the disclosure, the plurality of different machine learning of the data learning unit can be performed by selecting respective different regression model algorithms, but the kind of machine learning is not limited thereto.

[0052] More specifically, in one embodiment of the disclosure, the regression model algorithm can be one or more selected from a decision tree, a support vector machine (SVM), a random forest, a partial least squares regression, a quantile regression, a gradient boosting machine, a deep neural network, and a generalized linear / nonlinear regression, but the disclosure is not limited thereto.

[0053] Since the technology related to machine learning is well known in the technical field to which the disclosure pertains, a detailed description of a specific learning algorithm will be omitted.

[0054] In the present specification, the correlation of the capacity (Y) of the battery with the capacity factors (X1, X2,..., X n ) is derived in the form of an equation or a rule through machine learning. The capacity factor refers to values including voltage, current, capacity, impedance, temperature, etc. measured, collected, and stored during charging, discharging, and standing of the battery, which affect the capacity of the battery.

[0055] Specifically, the embodiment derived for the capacity (Y) of the battery obtained through machine learning can be represented by an equation Y = f(X1, X2,..., X n ). Here, f(X1, X2,..., X n ) refers to a functional form of the capacity factors (X1, X2,..., X n ), and includes a combination of all mathematical functions that derive a value equal to or approximately equal to the capacity (Y) of the battery. Here, the combination of the mathematical functions of the capacity factors (X1, X2,..., X n ) that most accurately predict the capacity (Y) of the battery is obtained as a result of machine learning on data in which the capacity and the capacity factors of the battery are measured, collected, and stored. That is, in the process of performing machine learning, the combination of the mathematical functions of the capacity factors that minimizes the deviation from the actual value of the capacity (Y) is obtained.

[0056] As another embodiment, the result obtained through machine learning can be represented as an IF-THEN rule. Here, the IF-THEN rule refers to if the capacity factors satisfy a plurality of specific conditions IF{(X1, X2,..., X n ), then the capacity (Y) of the battery has a certain specific value or a value within a certain range (THEN Y = yi). Here, the plurality of specific conditions IF{(X1, X2,..., X n} refers to each individual capacity factor (X1, X2, ..., Xn) or a function combination consisting of several capacity factors having a specific value or a value within a specific range. Multiple specific conditions IF{(X1, X2, ..., Xn) represent the specific value or value within a specific range for each individual capacity factor or a function combination consisting of several capacity factors. n A condition can have a hierarchical structure between conditions. That is, some specific conditions can be applied after some other specific conditions have been applied.

[0057] Furthermore, when learning from data on battery capacity and capacity factor measurements, collection, and storage by applying multiple machine learning algorithms, some machine learning methods can learn without splitting the entire dataset, or by dividing the entire dataset into several parts.

[0058] When learning by dividing the entire data into several parts, it can be done by dividing the entire data into a training dataset and a test dataset. The training dataset is used to derive mathematical formulas or IF-THEN rules from the machine learning algorithm, and the test dataset is used to evaluate the mathematical formulas or IF-THEN rules.

[0059] Specifically, the following method can be applied to part or all of the capacity range of the corresponding battery: After initially dividing the entire learning data, consisting of battery capacity and capacity factor used for machine learning, into several parts, and then using each segment of the dataset to learn from each of the multiple machine learning algorithms to create the same number of battery capacity prediction models as the learning algorithms, when predicting capacity from newly input capacity factor data, an average value is obtained by statistically summing the estimated capacity values ​​derived from the application of multiple battery capacity prediction models, and this average value is determined as the final battery capacity.

[0060] Furthermore, the capacity values ​​of batteries placed on modules, groups, or trays, or the capacity values ​​of batteries designed to have the same rated capacity and manufactured under the same manufacturing conditions, are represented as a distribution with variance, standard deviation, upper and lower limits. Multiple machine learning models are applied symmetrically or asymmetrically based on the center (mean or median) of the capacity distribution.

[0061] For example, such as Figure 1 As shown, the first to third machine learning models can be applied equally from the center (mean or median) of the capacity distribution to the side where the capacity decreases and the side where the capacity increases, based on an integer or real multiple of the standard deviation of the capacity range. The first to third machine learning models can be applied in the same way, or different machine learning models can be applied to the side where the capacity decreases and the side where the capacity increases.

[0062] Specifically, first, after learning data and creating capacity prediction models for a plurality of total capacity intervals placed on a module, group, or tray using a plurality of machine learning algorithms, the accuracy of the predicted capacity values and the error reduction rate of each machine learning algorithm are evaluated during the data learning process, and thus the priority between the prediction models is determined by a method of selecting the prediction model with the highest performance (i.e., accuracy and error reduction rate) in each capacity interval as the optimal machine learning model in the corresponding capacity interval.

[0063] Next, the capacity values of each battery for each prediction model are predicted using the capacity factor measurement data newly measured and collected from the individual batteries arranged on the module, group, or tray and the capacity prediction models derived from the plurality of machine learning. Thereafter, the capacity values returned by the machine learning prediction models having the predetermined priority are determined as the final capacity values of the corresponding batteries according to the capacity intervals to which the predicted values correspond.

[0064] As one example, as shown in Figure 1 , first, the capacity prediction values derived by applying the first machine learning model are assigned as the capacity values of the corresponding battery in the capacity interval below a times the standard deviation from the center (mean or median) of the capacity distribution of the battery to the side of the decreasing capacity value and the side of the increasing capacity value. In addition, as shown in Figure 1 , the capacity prediction values derived by applying the second machine learning model are assigned as the capacity values of the corresponding battery in the capacity interval greater than a times the standard deviation and equal to or less than b times the standard deviation from the center (mean or median) of the capacity distribution outside the capacity range of the battery derived by the first machine learning model to the side of the decreasing capacity value and the side of the increasing capacity value.

[0065] Similarly, as shown in Figure 1 , learning can be performed by assigning the capacity prediction values derived by applying the third machine learning model as the capacity values of the corresponding battery in the capacity range outside the capacity range of the battery respectively derived by the first and second machine learning models and the capacity interval greater than b times the standard deviation and equal to or less than c times the standard deviation from the center (mean or median) of the capacity distribution to the side of the decreasing capacity value and the side of the increasing capacity value.

[0066] In the embodiment shown in Figure 1 , random forest, gradient boosting machine, and quantile regression can be applied as the first to third machine learning methods, respectively, but this merely illustrates the algorithms of the first to third machine learning methods. In addition, learning can be performed in the same manner by applying other machine learning algorithms.

[0067] Figure 1 The definitions of the symbols shown in

[0068] Di = input data

[0069] D o = output data

[0070] σ = standard deviation of capacity, a, b, c = integer or real number

[0071] D aσ = capacity value data within a times standard deviation (aσ) range from the center (mean or median) of the capacity distribution

[0072] D bσ = capacity value data within b times standard deviation (bσ) range from the center (mean or median) of the capacity distribution

[0073] D cσ = capacity value data within c times standard deviation (cσ) range from the center (mean or median) of the capacity distribution

[0074] That is, depending on various conditions such as the kind of battery set on a module, group, or tray, the manufacturing method of the battery, the structure of the battery, the operating conditions, etc., the combination of machine learning algorithms that show the best performance (accuracy and error reduction rate) for the capacity prediction value is different, and is not limited only to the cases presented in the above examples.

[0075] The application criteria of the predicted capacity value from the machine learning model follow the priority of the machine learning model and the capacity interval predetermined in the learning process. For example, for each capacity interval, the determination coefficient (R square, R 2 ), mean absolute error (MAE), root mean square error (RMSE), mean absolute percentage error (MAPE), etc. are calculated to determine the range of the capacity interval with the optimized value. For example, the integer or real multiple of the standard deviation of the capacity range where the value of the determination coefficient becomes the maximum value or the mean absolute error, the root mean square error, or the mean absolute percentage error becomes the minimum value, i.e., the capacity distribution is obtained, and the machine learning algorithm that exhibits the highest accuracy and error reduction rate within each capacity interval is preferentially applied.

[0076] When the battery capacity value predicted from the machine learning model deviates from the upper and lower limits of the rated capacity, the battery is diagnosed as defective.

[0077] In this specification, an "input unit" is an interface for receiving various types of necessary data. Specifically, in this specification, the input unit can be divided into a learning data input unit for receiving learning data and a measurement data input unit for receiving measurement data. More specifically, an "input unit" is an interface for measuring or collecting capacity factors under rated capacity conditions and transmitting the measured or collected capacity factor measurement data to a reference value storage unit or a data learning unit. There are no particular limitations on the methods by which the learning unit receives and transmits data.

[0078] In this specification, "data learning unit" is an interface used to perform machine learning using learning data input to the learning data input unit.

[0079] In this specification, "output unit" is an interface used to compute predictive data by reflecting the results of machine learning. There are no particular limitations on the method by which the output unit computes the data.

[0080] In one embodiment of this disclosure, the apparatus for measuring battery capacity may further include: a reference value storage unit for storing data measured under battery rated capacity conditions; and a capacity status diagnosis unit for comparing the output battery capacity prediction data with the results of the data measured under battery rated capacity conditions to determine the reliability of the prediction data, diagnose the battery capacity and status, and control the battery process based on the diagnosis results.

[0081] In this specification, the "reference value storage unit" is an interface used to store capacity factor measurement data measured or acquired under rated capacity conditions, calculate a capacity reference value using this capacity factor measurement data, and transmit this capacity reference value to the capacity status diagnostic unit. There are no particular limitations on the method by which the reference value storage unit stores data and transmits specific values.

[0082] All data input, transmitted, or computed to the interface of the measuring device according to this disclosure can be managed uniformly. Here, uniform management may include, for example, all actions of managing all data input, transmitted, or computed to the interface of the measuring device according to this disclosure via a specific host computer or server, calculating new values ​​from the managed data, or re-inputting it as data to the input unit.

[0083] In this specification, the "capacity status diagnostic unit" is an interface used to control the battery process based on the diagnostic results obtained by comparing the capacity reference value received from the reference value storage unit with the capacity prediction value derived from data learning, thereby determining the reliability of the prediction data and diagnosing the battery's capacity and status.

[0084] This structure is like Figure 2 As shown, and except for the output unit below Figure 2In addition to the interfaces shown in the middle, there can be additional interfaces as necessary.

[0085] In one embodiment of the present disclosure, the device for measuring battery capacity can compare the output battery capacity prediction data with the actual capacity data results of the battery using a box plot.

[0086] In one embodiment of the present disclosure, determining the reliability of the battery capacity prediction data in the device for measuring battery capacity can be determining a coefficient of determination (R-squared, R 2 ), mean absolute error (MAE), root mean square error (RMSE), or mean absolute percentage error (MAPE), but the present disclosure is not limited thereto.

[0087] Since the art described in the present disclosure is widely known about the method of comparing data and determining reliability, a detailed description thereof will be omitted.

[0088] In one embodiment of the present disclosure, the battery capacity range of the battery capacity distribution in the device for measuring battery capacity can be determined as an integer multiple or a real multiple based on the standard deviation (σ) of the center of the capacity distribution.

[0089] The center of the capacity distribution can refer to the mean (Mean) or median (Median) of the capacity distribution.

[0090] In one embodiment of the present disclosure, the battery capacity range of the battery capacity distribution in the device for measuring battery capacity can be determined as an integer multiple or a real multiple based on the standard deviation (σ) of the mean (Mean) or median (Median) of the capacity distribution.

[0091] For example, in the present embodiment, when the collection regression method is used in the capacity interval of 1.5 times the standard deviation from the center of the capacity distribution, and the quantile regression method is used in the capacity interval greater than 1.5 times the standard deviation, the accuracy rate of predicting the capacity can be the maximum.

[0092] In an embodiment of the present disclosure, a battery management system (BMS) device including the device for measuring battery capacity according to the present disclosure can be provided. In other words, in an embodiment of the present disclosure, the device for measuring battery capacity can be used in a battery management system (BMS) device.

[0093] In an embodiment of the present disclosure, a battery management system device including a capacity measurement device including a learning data input unit, a measurement data input unit, a data learning unit, and an output unit can be provided.

[0094] In the present specification, the "battery management system (BMS) device" refers to all types of interfaces including a battery management system.

[0095] In one embodiment of the present disclosure, a mobile device can be provided including the management system device according to the present application.

[0096] In the present specification, the term "mobile device" refers to a device that can move by itself or can be easily carried by a user, and examples of the mobile device can include an electric vehicle, a mobile device, etc.

[0097] In one embodiment of the present disclosure, a battery management system device can be provided in which at least one of a learning data input unit, a measurement data input unit, a data learning unit, a reference value storage unit, and a capacity state diagnosis unit is remotely controlled.

[0098] In one embodiment of the present disclosure, a battery management system device can be provided in which two or more of a learning data input unit, a measurement data input unit, a data learning unit, a reference value storage unit, and a capacity state diagnosis unit are remotely controlled.

[0099] In one embodiment of the present disclosure, a battery management system device can be provided in which all of a learning data input unit, a measurement data input unit, a data learning unit, a reference value storage unit, and a capacity state diagnosis unit are remotely controlled.

[0100] In the present specification, the meaning of "being remotely controlled" refers to interfaces of input units, learning units, output units, etc. being located outside the battery management system device, so that their functions are performed while data and signals are transmitted or received between the interfaces through communication. For example, the method of remote control can include a method of managing the execution of their functions while data and signals are transmitted or received between the interfaces through communication by placing some interfaces in a cloud server, but the present disclosure is not limited thereto, and any method capable of performing their functions outside the battery management system device can be applied to the configuration of the present disclosure.

[0101] When some or all of the interfaces of the device are remotely controlled, the weight of the battery management system device can be reduced, so that it is easily applied to a mobile device, and using a specific host computer or a cloud server, so that it is easily managed integrally data, etc. generated in the process of using the device.

[0102] Further, when some of the interfaces of the device are remotely controlled, the cost related to computer hardware (H / W) installed in a mobile device can be reduced by reducing the specifications required for memory, computation, information processing, etc. for data storage related to the computer hardware (H / W) and simplifying the configuration.

[0103] Further, in one embodiment of the present disclosure, at least one of the learning data input unit; the measurement data input unit; the data learning unit; the output unit; the reference value storage unit; and the capacity state diagnosis unit of the battery management system device can be embedded in the mobile device.

[0104] Further, in one embodiment of the present disclosure, two or more of the learning data input unit; the measurement data input unit; the data learning unit; the output unit; the reference value storage unit; and the capacity state diagnosis unit of the battery management system device can be embedded in the mobile device.

[0105] Further, in one embodiment of the present disclosure, all of the learning data input unit; the measurement data input unit; the data learning unit; the output unit; the reference value storage unit; and the capacity state diagnosis unit of the battery management system device can be embedded in the mobile device.

[0106] In one embodiment of the present disclosure, a mobile device can be provided, in which at least one of the learning data input unit; the measurement data input unit; the data learning unit; the output unit; the reference value storage unit; and the capacity state diagnosis unit is embedded in the mobile device.

[0107] In one embodiment of the present disclosure, a mobile device can be provided, in which two or more of the learning data input unit; the measurement data input unit; the data learning unit; the output unit; the reference value storage unit; and the capacity state diagnosis unit of the battery management system device are embedded in the mobile device.

[0108] In one embodiment of the present disclosure, a mobile device can be provided, in which all of the learning data input unit; the measurement data input unit; the data learning unit; the reference value storage unit; and the capacity state diagnosis unit are embedded in the mobile device.

[0109] In the present specification, the term "embedded in the mobile device" means that the interface of the above-mentioned input unit, learning unit, output unit, etc. corresponds to one of the components of the mobile device.

[0110] When the device is partially or entirely embedded in the mobile device, it has the advantage that no security problems due to communication problems occur.

[0111] In one embodiment of the present disclosure, the device for measuring battery capacity constituting the battery management system device can further include the above-mentioned reference value storage unit; and a capacity state diagnosis unit, the reference value storage unit and the capacity state diagnosis unit can be partially embedded in a remote control or mobile device. If there is an additional interface available, each interface can be embedded in a remote control or mobile device.

[0112] More specifically, when the battery management system is installed on an electric vehicle, a mobile device, etc. and used as a power source, the battery management system refers to a system that performs capacity matching and cell balancing, controls charging or discharging of a battery, and controls and manages the overall state of a battery, such as the remaining capacity of a battery, a battery failure, etc. The battery management system (BMS) can be applied to one or more batteries. That is, it is generally applied to a plurality of batteries, but can also be applied to one battery, and the battery management system can be applied to each battery individually.

[0113] The data generated by the battery management system device can also be integrally managed as described above.

[0114] When the device for measuring battery capacity according to the present disclosure is applied to a battery management system device, the accuracy of battery capacity measurement can be improved, and thus the accuracy of diagnosing the state of a battery and predicting the life of a battery can be correspondingly improved. That is, since one or more batteries are installed on an electric vehicle, a mobile device, etc. and used as a power source, when overall battery management such as capacity matching, cell balancing, etc. is performed, the battery can be more accurately and effectively managed by the battery management system.

[0115] In one embodiment of the present disclosure, a method for measuring battery capacity is provided, the method including the steps of: inputting battery capacity factor learning data in a charging and discharging process of a battery selected as a learning target for a certain time; deriving a battery capacity distribution from the input battery capacity factor learning data; performing a plurality of different machine learnings for each battery capacity range of the battery capacity distribution derived from the learning data, respectively; inputting battery capacity factor measurement data selected in a charging and discharging process of a battery selected as a prediction target for a certain time; calculating battery capacity prediction data from the input battery capacity factor measurement data; and outputting battery capacity prediction data calculated for each battery capacity range of the battery capacity distribution derived from the learning data by the results of the plurality of machine learnings, respectively.

[0116] In one embodiment of the present disclosure, the battery capacity factor learning data for the method of measuring battery capacity can include battery charge voltage, battery discharge voltage, battery open circuit voltage (OCV), battery charge current, battery discharge current, battery charge capacity, battery discharge capacity, battery impedance, and battery temperature measured during battery charging, discharging, and resting by capacity measurement values corresponding to the rated capacity of a single battery selected as a learning target, but the present disclosure is not limited thereto, and any factor that can affect battery capacity can be included herein.

[0117] In one embodiment of the present disclosure, the battery capacity factor learning data for the method of measuring battery capacity can include battery charge voltage, battery discharge voltage, battery open circuit voltage (OCV), battery charge current, battery discharge current, battery charge capacity, battery discharge capacity, battery impedance, and battery temperature measured during battery charging, discharging, and resting by capacity measurement values corresponding to the rated capacity of a single battery selected as a learning target.

[0118] In one embodiment of the present disclosure, the battery capacity factor learning data for the method of measuring battery capacity can include battery charge voltage, battery discharge voltage, battery open circuit voltage (OCV), battery charge current, battery discharge current, battery charge capacity, battery discharge capacity, battery impedance, and battery temperature measured during battery charging, discharging, and resting by capacity measurement values corresponding to the rated capacity of a single battery selected as a learning target.

[0119] In one embodiment of the present disclosure, the battery capacity factor learning data for the method of measuring battery capacity can include battery charge voltage, battery discharge voltage, battery open circuit voltage (OCV), battery charge current, battery discharge current, battery charge capacity, battery discharge capacity, battery impedance, and battery temperature measured during battery charging, discharging, and resting by capacity measurement values corresponding to the rated capacity of a single battery selected as a learning target.

[0120] In one embodiment of the present disclosure, the step of machine learning on the input learning data can be performed by one or more selected from a decision tree, a support vector machine (SVM), a random forest, a partial least squares regression, a quantile regression, a gradient boosting machine, a deep neural network, and a generalized linear / nonlinear regression, but the present disclosure is not limited thereto. For example, in the present embodiment, the accuracy rate of predicting the capacity can become the maximum when the ensemble regression method is used in a capacity interval of 1.5 times the standard deviation from the center of the capacity distribution, and the quantile regression method is used in a capacity interval greater than 1.5 times the standard deviation.

[0121] In one embodiment of the present disclosure, the method for measuring battery capacity can further include the steps of storing actual capacity data of the battery; and comparing the output battery capacity prediction data with the actual capacity data of the battery to determine reliability of the battery capacity prediction data.

[0122] In one embodiment of the present disclosure, in the method for measuring battery capacity, the output battery capacity prediction data can be compared with the actual capacity data of the battery using a box plot, but the present disclosure is not limited thereto.

[0123] In one embodiment of the present disclosure, the step of determining reliability of the battery capacity prediction data can use actual standard capacity distribution of the battery stored in the reference value storage unit, a coefficient of determination (R-square, R 2 ), a mean absolute error (MAE), a root mean square error (RMSE), or a mean absolute percentage error (MAPE), but the present disclosure is not limited thereto.

[0124] In one embodiment of the present disclosure, the battery capacity range of the battery capacity distribution in the method for measuring battery capacity can be determined as an integer multiple or a real multiple based on a standard deviation (σ) of a center of the capacity distribution.

[0125] The center of the capacity distribution can refer to a mean (Mean) or a median (Median) of the capacity distribution.

[0126] In one embodiment of the present disclosure, the battery capacity range of the battery capacity distribution in the method for measuring battery capacity can be determined as an integer multiple or a real multiple based on a standard deviation (σ) of a mean (Mean) or a median (Median) of the capacity distribution.

[0127] In an embodiment of the present disclosure, a plurality of machine learning can be connected in parallel.

[0128] In an embodiment of the present disclosure, the method for measuring battery capacity can be a method used in a battery management system (BMS). That is, in an embodiment of the present disclosure, the battery management system can perform the functions of the battery management system described above by using the method for measuring battery capacity according to the present disclosure.

[0129] In this case, as described above, since one or more batteries are installed on an electric vehicle, a mobile device, etc. and used as a power source, when overall battery management such as capacity matching, cell balancing, etc. is performed, the battery can be more accurately and effectively managed by the battery management system.

[0130] In the present specification, the description applied to the device for measuring battery capacity according to the embodiments of the present disclosure can also be applied to the method for measuring battery capacity according to the embodiments of the present disclosure.

[0131] The device and method for measuring battery capacity according to the present application apply a method capable of maximizing the accuracy among a plurality of machine learning methods to battery capacity factor learning data of each capacity interval, so that the accuracy and precision of battery capacity prediction can be improved. As such, the efficiency of battery state diagnosis and quality control can be improved, and the process optimization and production efficiency can ultimately be improved.

[0132] When a plurality of batteries in the form of modules, groups, and trays are mounted on an electric vehicle, a mobile device, etc., and used as a power source, the device and method for measuring battery capacity according to the embodiments of the present application can improve the capacity measurement accuracy of individual batteries when capacity matching and cell balancing are performed, and as a result, the lifespan of batteries in the form of modules, groups, and trays can be improved.

[0133] The embodiments of the present disclosure provide a computer program stored in a recording medium for executing the method for measuring battery capacity according to the present disclosure. The above description of the method for measuring battery capacity can be applied in the same manner, except that each step of the method for measuring battery capacity is stored in the recording medium in the form of a computer program.

[0134] The process of the method for measuring battery capacity will be described in more detail below.

[0135] <EMBODIMENT>

[0136] After placing one or more batteries to be learned on a module, a group, and a tray, and measuring and collecting battery capacity factors such as charging voltage, discharging voltage, open circuit voltage (OCV), charging current, discharging current, charging capacity, discharging capacity, impedance, and temperature of the placed batteries while performing a charging and discharging process of the battery corresponding to a capacity measurement value of a rated capacity, the values are stored in a storage medium as capacity factor learning data.

[0137] Thereafter, a capacity distribution is derived from the measured and collected capacity factor learning data, machine learning is performed using the first to third machine learning models for each battery capacity range of the capacity distribution derived from the learning data, and three battery capacity prediction models are derived therefrom.

[0138] Specifically, the capacity prediction value derived by applying the first machine learning model is assigned as the capacity value of the corresponding battery within the range of 1.5 times the standard deviation based on the center of the capacity distribution, and the capacity prediction value derived by the second machine learning model is assigned as the capacity value of the corresponding battery with respect to the battery capacity range derived by the first machine learning model, i.e., in the case of outside the range of 1.5 times the standard deviation based on the center of the capacity distribution. More specifically, the battery capacity range to which the second machine learning model is applied refers to the range of 1.5 to 2 times the standard deviation.

[0139] Similarly, when the battery capacity derived by applying the first and second machine learning models exceeds the range, i.e., when it deviates by two times the standard deviation, the capacity prediction value derived by applying the third machine learning model is assigned as the capacity value of the corresponding battery.

[0140] More specifically, the random forest (first machine learning model) is applied within the range of 1.5 times the standard deviation based on the center of the capacity distribution, and the gradient boosting machine (second machine learning model) algorithm is applied within the range of 1.5 to 2 times the standard deviation based on the center of the capacity distribution.

[0141] Finally, the quantile regression (third machine learning model) algorithm is applied with respect to the range deviating by two times the standard deviation based on the center of the capacity distribution.

[0142] In addition, both bagging and boosting methods are used in this process.

[0143] At this time, the accuracy is maximized when the bagging and boosting algorithms are applied within the capacity range from the center of the distribution to 1.5 to 2 times the standard deviation, and the accuracy is maximized when the quantile regression method is applied in the capacity range outside thereof.

[0144] Thereafter, after placing one or more batteries for which the capacity is to be predicted on a module, a pack, and a tray, and measuring and collecting battery capacity factors such as a charging voltage, a discharging voltage, an open circuit voltage (OCV), a charging current, a discharging current, a charging capacity, a discharging capacity, an impedance, and a temperature of the placed batteries while performing a charging and discharging process of the batteries, the values are stored in a storage medium as capacity factor measurement data learning data.

[0145] Thereafter, battery capacity prediction data is calculated by applying the three derived battery capacity measurement models to the capacity factor measurement data.

[0146] Next, a battery capacity distribution is derived to predict the battery capacity by outputting the calculated battery capacity prediction data for each capacity range of the capacity distribution derived from the capacity factor learning data.

[0147] <Comparative Example>

[0148] In addition, battery capacity prediction data was derived in the same manner except that a linear regression equation was applied as a single machine learning algorithm (Comparative Example).

[0149] The derivation results according to the embodiment and the comparative example were displayed through a capacity distribution comparison graph of Figure 3 and a box plot of Figure 4 .

[0150] The meanings of (a) to (d) in Figure 3 and Figure 4 are as follows.

[0151] (a) Nominal capacity

[0152] (b) Capacity calculated through a single machine learning (ML) method

[0153] (c) Capacity calculated through multiple machine learning (ML) methods (Embodiment of the disclosure)

[0154] (d) Capacity calculated from a linear regression equation

[0155] It can be visually confirmed from the results of Figure 3 and Figure 4 that the difference between the measured data and the predicted data is not large and the accuracy is excellent in the case of the device and method for measuring battery capacity according to the embodiment of the present application.

[0156] Finally, the actual capacity data of the battery was compared with the output battery capacity prediction data to determine the reliability of the battery capacity prediction data by deriving R 2 .

[0157] In addition, battery capacity prediction data was derived in the same manner except that a single machine learning algorithm was applied to derive R 2 to determine the reliability thereof.

[0158] As a result, it was confirmed that the device and method for measuring battery capacity according to the embodiment of the present application improved R 2 by 20% or more compared to the case in which battery capacity was measured in the same manner except that a single machine learning method was applied.

Claims

1. A device for measuring battery capacity, the device comprising: The learning data input unit is used to receive battery capacity factor learning data measured during a specific period of charging and discharging of a single battery selected as the learning target. The measurement data input unit is used to receive battery capacity factor measurement data selected during the charging and discharging process of the battery selected as the prediction target for a specific time. The data learning unit is used to derive the battery capacity distribution from the battery capacity factor learning data of the input data to the learning data unit, and to perform multiple different machine learning operations for each battery capacity range of the battery capacity distribution derived from the learning data. The output unit is used to calculate the capacity prediction data of the battery selected as the prediction target from the input battery capacity factor measurement data through the results of multiple different machine learning, and output the battery capacity prediction data calculated for each battery capacity range of the battery capacity distribution derived from the learning data. Reference value storage unit, used to store data measured under battery rated capacity conditions; as well as The capacity status diagnostic unit is used to compare the output battery capacity prediction data with the results of data measured under the battery's rated capacity conditions to determine the reliability of the battery capacity prediction data, diagnose the battery's capacity and status, and control the battery process based on the diagnostic results.

2. The apparatus according to claim 1, wherein, The battery capacity factor learning data includes battery charging capacity and battery discharging capacity measured during charging, discharging, and resting of the battery by capacity measurements corresponding to the rated capacity of the individual battery selected as the learning target, and further includes one or more of battery charging voltage, battery discharging voltage, battery open circuit voltage (OCV), battery charging current, battery discharging current, battery impedance, and battery temperature.

3. The apparatus according to claim 1, wherein, Battery capacity factor measurement data includes one or more of the following: battery charging voltage, battery discharging voltage, battery open circuit voltage (OCV), battery charging current, battery discharging current, battery charging capacity, battery discharging capacity, battery impedance, and battery temperature, measured during charging, discharging, and resting of the battery selected as the prediction target.

4. The apparatus according to claim 1, wherein, The data learning unit performs multiple different machine learning operations by selecting its own different regression model algorithms.

5. The apparatus according to claim 4, wherein, The regression model algorithm is selected from one or more of the following: decision tree, support vector machine (SVM), random forest, partial least squares regression, quantile regression, gradient boosting machine, deep neural network, and generalized linear / nonlinear regression.

6. The apparatus according to claim 1, wherein, Using the capacity distribution measured under battery rated capacity conditions and stored in the reference value storage unit, the coefficient of determination (R², R...) 2 The reliability of battery capacity prediction data is determined by mean absolute error (MAE), root mean square error (RMSE), or mean absolute percentage error (MAPE).

7. The apparatus according to claim 1, wherein, The battery capacity range is determined by multiples or real multiples of the standard deviation (σ) of the mean or median of the capacity distribution.

8. A method for measuring battery capacity, the method comprising the following steps: Input battery capacity factor learning data during a specific charging and discharging process of the battery selected as the learning target; Derive the battery capacity distribution from the input battery capacity factor learning data; For each battery capacity range of the battery capacity distribution derived from the learning data, perform multiple different machine learning operations; Input the battery capacity factor measurement data selected during a specific period of charging and discharging of the battery selected as the prediction target; Based on the results of multiple machine learning methods, the capacity prediction data of the selected battery is calculated from the input battery capacity factor measurement data. Output battery capacity prediction data calculated for each battery capacity range of the battery capacity distribution derived from the learning data; Store and update battery capacity data measured under rated battery capacity conditions; as well as The output battery capacity prediction data is compared with the battery capacity data measured under the battery's rated capacity conditions to determine the reliability of the battery capacity prediction data.

9. The method according to claim 8, wherein, The battery capacity factor learning data includes battery charging capacity and battery discharging capacity measured during charging, discharging, and resting of the battery by capacity measurements corresponding to the rated capacity of the individual battery selected as the learning target, and further includes one or more of battery charging voltage, battery discharging voltage, battery open circuit voltage (OCV), battery charging current, battery discharging current, battery impedance, and battery temperature.

10. The method according to claim 8, wherein, Battery capacity factor measurement data includes one or more of the following: battery charging voltage, battery discharging voltage, battery open circuit voltage (OCV), battery charging current, battery discharging current, battery charging capacity, battery discharging capacity, battery impedance, and battery temperature, measured during charging, discharging, and resting of the battery selected as the prediction target.

11. The method according to claim 8, wherein, The input learning data is subjected to multiple different machine learning steps by selecting different regression model algorithms.

12. The method according to claim 11, wherein, The different regression model algorithms are selected from one or more of decision trees, support vector machines (SVM), random forests, partial least squares regression, quantile regression, gradient boosting machines, deep neural networks, and generalized linear / nonlinear regression.

13. The method according to claim 8, wherein, The steps to determine the reliability of battery capacity prediction data are to use the capacity distribution measured under battery rated capacity conditions stored in the reference value storage unit, and the coefficient of determination (R², R₀²). 2 The mean absolute error (MAE), root mean square error (RMSE), or mean absolute percentage error (MAPE) are used to measure the error.

14. The method according to claim 8, wherein, The battery capacity range is determined by multiples or real multiples of the standard deviation (σ) of the mean or median of the capacity distribution.

15. A battery management system (BMS) device comprising means for measuring battery capacity according to any one of claims 1 to 7.

16. The battery management system (BMS) device according to claim 15, wherein, Learning data input unit; Measurement data input unit; Data learning unit; Output unit; Reference value storage unit; At least one of the capacity status diagnostic units is remotely controlled.

17. A mobile device comprising a battery management system (BMS) device according to claim 15.

18. The mobile device according to claim 17, wherein, A learning data input unit; a measurement data input unit; a data learning unit for embedding a battery management system in a mobile device; Output unit; Reference value storage unit; and at least one of the capacity status diagnostic units.

19. A computer program stored in a recording medium for performing a method for measuring battery capacity according to any one of claims 8 to 14.

Images