Soh prediction apparatus and method

Abstract

An SOH prediction device according to one embodiment of the present disclosure includes a data acquisition unit configured to acquire an actual observation histogram representing a correspondence between voltage values, current values, temperature values, and frequency values of a battery during an actual observation period; and a control unit configured to generate first to nth histograms corresponding to first to nth periods after the actual observation period based on the actual observation histogram and a reference temperature dataset, and predict a target SOH representing an SOH of the battery at an end point of the nth period based on the first to nth histograms and an actual observation SOH representing the SOH of the battery at the end point of the actual observation period according to a pre-set deterioration prediction model, where n is a natural number greater than or equal to 2.

Description

Technical Field

[0001] This disclosure relates to an apparatus and method for predicting the state of harmonics (SOH) of a battery using a pre-learned degradation prediction model.

[0002] This application is based on and claims priority to Korean Patent Application No. 10-2025-0001430, filed with the Korean Intellectual Property Office on January 6, 2025, the disclosure of which is incorporated herein by reference in its entirety. Background Technology

[0003] In recent years, the demand for portable electronic products such as laptops, cameras, and mobile phones has increased dramatically, and electric vehicles, energy storage batteries, robots, and satellites have also seen significant development. Therefore, high-performance batteries that allow for repeated charging and discharging are being actively researched.

[0004] Currently available batteries include nickel-cadmium (NiCd), nickel-metal hydride (NiMH), nickel-zinc (NiZn), and lithium-ion batteries. Among them, lithium-ion batteries are particularly noteworthy because, compared to nickel-based batteries, they have virtually no memory effect and also exhibit extremely low self-discharge rates and high energy density.

[0005] Battery State of Health (SOH) prediction technology is crucial for assessing battery performance and remaining life, as well as ensuring battery stability and efficiency. It is widely used in diverse applications, including electric vehicles, energy storage systems (ESS), and portable electronic devices, and has been established as a core function of battery management systems (BMS).

[0006] Existing SOH prediction technologies rely on physical models based on the battery's physical and chemical properties. These physical models offer the advantage of high prediction accuracy by mathematically modeling electrochemical reactions, thermal behavior, and physical changes within the battery. However, physical models require complex calculations, resulting in long computation times and making them difficult to apply to environments requiring real-time processing, such as automotive BMS or cloud-based BMS. Furthermore, physical models are optimized for specific battery configurations or usage conditions, limiting their versatility across diverse environments.

[0007] To overcome these limitations, data-driven State of Health (SOH) prediction models are gaining attention. Data-driven models predict battery state by learning from battery measurements (e.g., voltage, current, temperature, etc.), offering advantages such as faster computation and simpler implementation compared to physical models. However, existing data-driven models typically provide lower prediction accuracy than physical models, thus limiting their practical application.

[0008] Therefore, it is necessary to develop a new SOH prediction model that combines the advantages of high prediction accuracy as a physical model with the advantages of fast computation speed as a data-based model. Summary of the Invention

[0009] Technical issues

[0010] This disclosure is designed to address the problems of related technologies, and therefore aims to provide a SOH prediction apparatus and method capable of predicting the target SOH at a specific point in the future based on past actual observation data and a pre-learned degradation prediction model.

[0011] Other objects and advantages of this disclosure will be understood from the following detailed description and will become even more apparent from the exemplary embodiments thereof. Furthermore, it will be readily understood that the objects and advantages of this disclosure can be achieved by the means and combinations thereof shown in the appended claims.

[0012] Technical solution

[0013] A SOH prediction apparatus according to one aspect of this disclosure may include: a data acquisition unit configured to acquire a real-observation histogram representing the correspondence between voltage, current, temperature, and frequency values ​​of the battery during an actual observation period; and a control unit configured to generate first to nth histograms corresponding to first to nth time periods following the actual observation period based on the real-observation histograms and a reference temperature dataset, and to predict a target SOH representing the SOH of the battery at the end of the nth time period based on a pre-set degradation prediction model, according to the first to nth histograms and the real-observed SOH representing the SOH of the battery at the end of the actual observation period. n is a natural number greater than or equal to 2.

[0014] The reference temperature dataset may include the first to nth reference temperature values ​​corresponding to the first to nth time periods.

[0015] The control unit can be configured to generate the first to nth histograms based on the comparison between the representative temperature value of the actual observed histogram and each of the first to nth reference temperature values.

[0016] The control unit can be configured to generate a reference temperature dataset based on multiple temperature datasets collected from multiple other batteries during past data collection periods with a time length greater than or equal to the total time length of the first to nth time periods.

[0017] The control unit can be configured to determine a j-th temperature correction value, which represents the difference between the temperature value and the j-th reference temperature among the first to n-th reference temperature values. j is a natural number less than or equal to n.

[0018] The control unit can be configured to correct each temperature value of the actual observed histogram based on the j-th temperature correction value to generate the j-th histogram among the first to nth histograms.

[0019] The degradation prediction model can be trained using multiple learning datasets collected from multiple other batteries over multiple reference periods.

[0020] Each of the multiple learning datasets may include a reference histogram representing the correspondence between voltage, current, temperature, and frequency values ​​of other batteries during a reference period prior to the actual observation period, as well as SOH of other batteries at the beginning of the reference period as features, and SOH of other batteries at the end of the reference period as labels.

[0021] The length of each of the multiple reference time periods can be equal to the total length of the first to nth time periods.

[0022] The duration of each of the first to nth time periods can be equal to the duration of the actual observation period.

[0023] The degradation prediction model can be configured as a CNN (convolutional neural network) based model.

[0024] The control unit can be configured to predict the battery degradation rate during the first to nth time periods based on the actual observed SOH and the target SOH.

[0025] According to another aspect of this disclosure, the battery pack may include the SOH prediction device.

[0026] According to another aspect of this disclosure, a vehicle may include the SOH prediction device.

[0027] According to another aspect of this disclosure, the server may include the SOH prediction device.

[0028] A SOH prediction method according to another aspect of this disclosure may include: obtaining a real-observation histogram representing the correspondence between voltage, current, temperature, and frequency values ​​of the battery during an actual observation period; generating first to nth histograms corresponding to first to nth time periods after the actual observation period based on the real-observation histograms and a reference temperature dataset; and predicting a target SOH representing the SOH of the battery at the end of the nth time period based on a pre-set degradation prediction model, according to the first to nth histograms and the real-observed SOH representing the SOH of the battery at the end of the actual observation period. n is a natural number greater than or equal to 2.

[0029] Beneficial effects

[0030] According to one aspect of this disclosure, based on past actual observation data and a pre-learned degradation prediction model, the target SOH at a specific point in the future can be predicted with high confidence.

[0031] Furthermore, according to one aspect of this disclosure, the state of harmonics (SOH) of a battery can be predicted more accurately and reliably by using a 3D-CNN-based degradation prediction model.

[0032] The effects of this disclosure are not limited to those described above, and those skilled in the art will clearly understand from the description of the claims other unmentioned effects. Attached Figure Description

[0033] The accompanying drawings illustrate preferred embodiments of the present disclosure and, together with the foregoing disclosure, are intended to provide a further understanding of the technical features of the present disclosure; therefore, the present disclosure should not be construed as being limited to the drawings.

[0034] Figure 1 This is a schematic diagram illustrating an SOH prediction device according to an embodiment of the present disclosure.

[0035] Figure 2 This is a diagram that schematically illustrates an example of an actual observation histogram.

[0036] Figure 3 It is a diagram used as a reference to illustrate the actual observation period and the first to nth periods.

[0037] Figure 4 and Figure 5 This is a diagram schematically illustrating the structure of a degradation prediction model.

[0038] Figure 6 This is a schematic diagram illustrating a battery pack according to another embodiment of the present disclosure.

[0039] Figure 7 This is a schematic diagram illustrating a vehicle according to yet another embodiment of the present disclosure.

[0040] Figure 8 This is a schematic diagram illustrating a server according to yet another embodiment of the present disclosure.

[0041] Figure 9 This is a diagram schematically illustrating a SOH prediction method according to yet another embodiment of the present disclosure.

[0042] Figure 10 It is shown schematically. Figure 9 The diagram shows the sub-steps of step S920. Detailed Implementation

[0043] It should be understood that the terms used in the specification and appended claims should not be construed as limited to their general and dictionary meanings, but rather as being interpreted based on their meanings and concepts corresponding to the technical aspects of this disclosure, on the basis of the principle that the inventors are allowed to define the terms appropriately for the best interpretation.

[0044] Therefore, the description presented herein is merely a preferred example for illustrative purposes only and is not intended to limit the scope of this disclosure. It should be understood that other equivalents and modifications may be made thereto without departing from the scope of this disclosure.

[0045] Furthermore, in describing this disclosure, a detailed description of a relevant known element or function is omitted herein when it is considered that such a detailed description would obscure the key subject matter of the disclosure.

[0046] Ordinal terms such as “first” and “second” can be used to distinguish one element from another among various elements, but are not intended to limit these elements.

[0047] Throughout this specification, when a section is referred to as “comprising” or “including” any element, unless otherwise specified, it means that the section may further include other elements, without excluding other elements.

[0048] Furthermore, throughout the specification, when one part is referred to as being “connected” to another part, this is not limited to the case where they are “directly connected”, but also includes the case where they are “indirectly connected”, in which another element is inserted between them.

[0049] Preferred embodiments of the present disclosure will be described in detail below with reference to the accompanying drawings.

[0050] Figure 1 This is a schematic diagram illustrating an SOH prediction device 100 according to an embodiment of the present disclosure.

[0051] Reference Figure 1 The SOH prediction device 100 may include a data acquisition unit 110 and a control unit 120. The SOH prediction device 100 may also include a storage unit 130.

[0052] Data acquisition unit 110 can be configured to acquire data representing the battery (see...) Figure 6 The attached figure (11) shows the actual observation histogram of the correspondence between voltage, current, temperature and frequency values ​​during the actual observation period.

[0053] Here, a battery refers to a physically separable, individual cell with negative and positive terminals. For example, a lithium-ion battery or a lithium polymer battery can be considered a battery. Furthermore, the type of battery can be cylindrical, prismatic, or pouch-shaped. Additionally, a battery can refer to a battery bank, battery module, or battery pack containing multiple cells connected in series and / or parallel. In the following text, for ease of explanation, a battery will be interpreted as referring to a single, independent cell.

[0054] Figure 2 This is a diagram that schematically illustrates an example of an actual observation histogram.

[0055] The actual observation histogram can be the result of classifying VIT data points obtained during the actual observation period into one of the first to D VIT intervals. Here, each VIT data point can be a three-dimensional data value defined by voltage, current, and temperature values ​​indexed at the same measurement timing.

[0056] Each of the first to the D VIT intervals is a combination of any one of the first to the A voltage intervals (A is a natural number greater than or equal to 2), any one of the first to the B current intervals (B is a natural number greater than or equal to 2), and any one of the first to the C temperature intervals (C is a natural number greater than or equal to 2), such that D can be equal to the product of A, B, and C.

[0057] The first to Ath voltage ranges can be the entire voltage range that is permissible for the battery and is divided into A ranges; the first to Bth current ranges can be the entire current range that is permissible for the battery and is divided into B ranges; and the first to Cth temperature ranges can be the entire temperature range that is permissible for the battery and is divided into C ranges. For example, if the entire voltage range is 3.0 to 5.0 [V] and A = 4, then the first voltage range is 3.0 or higher to less than 3.5 [V], the second voltage range is 3.5 or higher to less than 4.0 [V], the third voltage range is 4.0 or higher to less than 4.5 [V], and the fourth voltage range is 4.5 or higher to less than 5.0 [V]. As another example, if the entire current range is -20 to +20 [A] and B = 4, then the first current range is -20 or higher to less than -10 [A], the second current range is -10 or higher to less than 0 [A], the third current range is 0 or higher to less than +10 [A], and the fourth current range is +10 or higher to less than +20 [A]. As another example, if the entire temperature range is -30 to +50 [°C] to C = 5, then the first temperature range is -30 or higher to less than -20 [°C], the second temperature range is -20 or higher to less than -5 [°C], the third temperature range is -5 or higher to less than +15 [°C], the fourth temperature range is +15 or higher to less than +35 [°C], and the fifth temperature range is +35 or higher to less than +50 [°C]. For reference, such as Figure 2 As shown, if A=4, B=4 and C=5, then D=80.

[0058] like Figure 2 As shown, the actual observation histogram can be represented in the form of a 3D matrix. Specifically, refer to... Figure 2 In actual observation, the row number of the histogram can represent the voltage range number, the column number can represent the current range number, and the depth number can represent the temperature range number.

[0059] Therefore, each coordinate of the 3D matrix representing the actual observation histogram can correspond to any one of the VIT intervals from the first to the Dth VIT interval. Furthermore, the value of each coordinate of the 3D matrix representing the actual observation histogram can represent the number (frequency value) of VIT data points classified into the VIT interval corresponding to that coordinate.

[0060] Reference Figure 2When 'a' is a natural number less than or equal to A, 'b' is a natural number less than or equal to B, and 'c' is a natural number less than or equal to C, the coordinates (a, b, c) can represent the VIT interval {AB(c-1) + (a-1)B + b}. For example, if the voltage, current, and temperature values ​​of a VIT data point are 3.2 [V], -18 [A], and -25 [℃], then this VIT data point can be classified into the first VIT interval corresponding to the coordinates (1, 1, 1). As another example, if the voltage, current, and temperature values ​​of another VIT data point are 4.6 [V], +17 [A], and +43 [℃], then this VIT data point can be classified into the 80th VIT interval corresponding to the coordinates (4, 4, 5).

[0061] Suppose that a total of E VIT data points are obtained during the actual observation period, and among them, F VIT data points are classified into the VIT interval {AB(c-1) + (a-1)B + b} corresponding to the coordinates (a, b, c). Then, F or F / E can be assigned as frequency values ​​to the coordinates (a, b, c) of the actual observation histogram.

[0062] The actual observation period refers to the time during which the battery's voltage, current, and temperature are measured. The actual observation period can be preset. For example, the actual observation period can be preset to the period from June 1, 2024 to June 30, 2024.

[0063] In one embodiment, the data acquisition unit 110 can directly measure the battery's voltage, current, and temperature. Specifically, the data acquisition unit 110 can measure the positive and negative terminal voltages via a pair of voltage sensing lines connected to the positive and negative terminals of the battery, respectively. Furthermore, the data acquisition unit 110 can measure the voltage across the two terminals of the battery based on the voltage difference between the measured positive and negative terminal voltages. The data acquisition unit 110 can be connected to the battery via a current measurement unit to measure the battery's current. For example, the current measurement unit can be a current sensor or a shunt resistor installed in the battery's charging / discharging path to measure the battery's current. Here, the battery's charging / discharging path can be a high-current path through which charging current is applied to the battery or discharging current is output from the battery. The data acquisition unit 110 can use a temperature sensor to measure the battery's temperature. That is, the data acquisition unit 110 can directly measure the battery's voltage, current, and temperature, and generate an actual observation histogram based on the measurement results.

[0064] In another embodiment, the data acquisition unit 110 can receive actual observation histograms from an external source. That is, the data acquisition unit 110 can receive actual observation histograms from an external source by connecting to an external source via wired and / or wireless communication. For example, the data acquisition unit 110 can receive actual observation histograms from an external source via CAN (Controller Area Network) communication or CAN-FD (CAN with Flexible Data Rate). As another example, the data acquisition unit 110 can receive actual observation histograms from an external source via Zigbee, Bluetooth, Wi-Fi, or a mobile communication network. Of course, the type of communication protocol is not particularly limited as long as it supports communication between the data acquisition unit 110 and the external source.

[0065] The data acquisition unit 110 can be connected to the control unit 120 via a wired and / or wireless connection to enable communication with the control unit 120. The data acquisition unit 110 can send the acquired actual observation histogram to the control unit 120. The control unit 120 can receive the actual observation histogram from the data acquisition unit 110.

[0066] The control unit 120 can be configured to generate first to nth histograms corresponding to the first to nth time periods after the actual observation period based on the actual observation histogram and the reference temperature dataset, where n is a natural number greater than or equal to 2.

[0067] The reference temperature dataset can include reference temperature values ​​for periods one through n corresponding to periods one through n. The period from the start of period one to the end of period n can be referred to as the target future period.

[0068] The first to nth time periods refer to consecutive time periods following the actual observation period. Each of the first to nth time periods can be set as a time unit (e.g., day, week, month) for predicting the target SOH based on the actual observation data. For example, if the time unit is "day", the start date of the first time period can be the day after the end date of the actual observation period, and the start date of the second time period can be the day after the end date of the first time period.

[0069] Specifically, the control unit 120 can generate the first to nth histograms by correcting the temperature values ​​of the actual observation histograms using a reference temperature dataset. At this time, assuming that the battery's usage pattern remains the same in each of the first to nth periods during the actual observation period, the patterns of the voltage and current values ​​measured during the actual observation period are also assumed to remain the same in each of the first to nth periods.

[0070] The control unit 120 can correct the temperature values ​​of the actual observed histogram by referring to the reference temperature values ​​of the first to nth time periods included in the reference temperature dataset, so as to reflect both temperature changes caused by battery use and temperature changes caused by the external environment. For example, the control unit 120 can correct the actual observed histogram by reflecting the external temperature changing over time. Specifically, the first to nth histograms reflecting high-temperature environments (summer) or low-temperature environments (winter) that have a significant impact on the battery's operating temperature and performance can be generated. In this way, the first to nth histograms reflecting the temperature conditions corresponding to each time period can be generated while maintaining the same voltage and current value patterns as the actual observed data.

[0071] A temperature correction process can be performed based on the first to nth reference temperature values ​​and the temperature values ​​included in the actual observation histogram. Through the temperature correction process, the control unit 120 can correct the temperature values ​​of the actual observation data by reflecting the battery's operating environment and external environmental factors. Based on the corrected data, the control unit 120 can generate first to nth histograms representing the correspondence between future voltage, current, temperature, and frequency values.

[0072] The control unit 120 can be configured to use a pre-set degradation prediction model to predict the target SOH representing the SOH of the battery at the end of the nth time period, based on the first to nth histograms and the actual observed SOH representing the SOH of the battery at the end of the actual observation period.

[0073] Figure 3 It is a reference diagram used to illustrate the actual observation period and the first to nth periods.

[0074] Reference Figure 3 The actual observation period refers to the interval during which the battery's state is monitored and data indicating changes in voltage, current, and temperature are collected. The State of Harm (SOH) calculated at the end of the actual observation period is defined as the actual observed SOH, and the histogram generated based on the data collected during the actual observation period is defined as the actual observed histogram.

[0075] The first to nth time periods represent the continuous intervals following the actual observation period. The SOH (State of Health) of the battery at the end of the first to nth time periods is defined as the target SOH. Specifically, the nth time period is the last interval in time, and the SOH of the battery at the end of the nth time period is defined as the target SOH. Furthermore, the first to nth histograms are respectively related to the first to nth time periods. That is, when j is a natural number less than or equal to n, the jth histogram is related to the jth time period. For example, the first histogram is a histogram showing the correspondence between the predicted voltage, current, temperature, and frequency values ​​for the first time period.

[0076] Specifically, the control unit 120 can predict the target SOH by inputting the first to nth histograms and the actual observed SOH into the degradation prediction model.

[0077] The degradation prediction model utilizes deep learning technology. Deep learning is a machine learning algorithm based on an artificial neural network consisting of multiple layers. Deep learning-based degradation prediction models can effectively learn the complex nonlinear relationships inherent in large amounts of data, thereby predicting the state of harmonics (SOH) of batteries with high reliability.

[0078] Specifically, the degradation prediction model can be a deep learning model trained by using a histogram representing the correspondence between voltage, current, temperature and frequency values, as well as the battery's state of harm (SOH) at a specific time point, as input to the learning dataset.

[0079] For example, a degradation prediction model can be a model trained by taking a reference histogram representing the correspondence between voltage, current, temperature and frequency values ​​during a reference period and the battery's state of harm (SOH) at the end of the reference period as input to a learning dataset.

[0080] That is, the SOH prediction device 100 can predict the target SOH at a specific point in the future with high reliability based on past actual observation data and a pre-learned degradation prediction model.

[0081] Meanwhile, the data acquisition unit 110 and / or control unit 120 included in the SOH prediction device 100 may optionally include processors, application-specific integrated circuits (ASICs), other chipsets, logic circuits, registers, communication modems, data processing devices, etc., known in the art to execute various control logics performed in this disclosure. Furthermore, when the control logic is implemented as software, the data acquisition unit 110 and / or control unit 120 can be implemented as a collection of program modules. In this case, the program modules can be stored in memory and executed by the data acquisition unit 110 and / or control unit 120. The memory can be internal or external to the data acquisition unit 110 and / or control unit 120, and can be connected to the data acquisition unit 110 and / or control unit 120 by various known means.

[0082] Furthermore, the SOH prediction device 100 may also include a storage unit 130. The storage unit 130 may store data required for the operation and function of each component of the SOH prediction device 100, data generated during the execution of operations or functions, etc. The storage unit 130 is not particularly limited, as long as it is a known information storage device capable of recording, erasing, updating, and retrieving data. As examples, the information storage device may include RAM (Random Access Memory), flash memory, ROM (Read-Only Memory), EEPROM (Electrically Erasable Programmable Read-Only Memory), registers, etc. Additionally, the storage unit 130 may store program code defining processes that can be executed by the data acquisition unit 110 and / or the control unit 120.

[0083] Specifically, storage unit 130 can store information required by control unit 120 to predict the target SOH of the battery. Furthermore, control unit 120 can access storage unit 130 to obtain the information required by control unit 120 to predict the target SOH of the battery. For example, the actual observation histogram acquired by data acquisition unit 110 can be stored in storage unit 130, and control unit 120 can access storage unit 130 to obtain the stored actual observation histogram. Degradation prediction models can be recorded in storage unit 130.

[0084] In one embodiment, the length of each of the first to nth time periods can be set to be equal to the length of the actual observation period. This allows the control unit 120 to generate the first to nth histograms as prediction data more consistently based on the actual observation histograms generated from the actual observation data. Specifically, consistency in time length can help improve the accuracy and reliability of the prediction data by standardizing the time standard during the data generation process.

[0085] The control unit 120 can be configured to generate the first to nth histograms based on the comparison results between the representative temperature value of the actual observed histogram and each of the first to nth reference temperature values.

[0086] Here, the representative temperature value refers to the representative value derived from the multiple temperature values ​​included in the actual observation histogram. For example, the representative value can be the mean, median, mode, or a value based on user-defined rules.

[0087] The first to nth reference temperature values ​​are values ​​used as standards for temperature correction and can be set according to predetermined standards. Furthermore, the first to nth reference temperature values ​​can constitute a reference temperature dataset.

[0088] For example, the control unit 120 can be configured to generate a reference temperature dataset based on multiple temperature datasets collected from multiple other batteries during past data collection periods with a time length greater than or equal to the total time length of the first to nth time periods.

[0089] Here, the multiple other batteries can include batteries of the same type as the target battery in the SOH prediction. These multiple other batteries can include batteries with different degrees of degradation, manufacturing environments, and usage conditions. If the target battery is included among the multiple other batteries, the target battery's temperature data can be directly used in the process of generating the reference temperature dataset, thereby generating a reference temperature dataset that more effectively reflects the target battery's usage conditions and environment. Conversely, if the target battery is not included among the multiple other batteries, the temperature data of the other batteries can be used to generate a more general reference temperature dataset.

[0090] Multiple temperature datasets refer to datasets that include temperature values ​​measured from multiple other batteries during past data collection periods. Each temperature dataset includes temperature values ​​measured from a single battery and can be configured to be temporally continuous data. Temperature datasets can include a variety of temperature values ​​reflecting battery usage environments and conditions during a specific time period. Each temperature value in multiple temperature datasets can be indexed using measurement timing information (e.g., month, day, hour, minute, and second).

[0091] The control unit 120 can classify multiple temperature values ​​included in multiple temperature datasets into datasets numbered from 1 to x according to predetermined rules. x is a natural number greater than or equal to n. For example, if the time unit for classification is "days," then x could be 365. In this case, the first dataset includes temperature values ​​measured on January 1st regardless of the year, the 32nd dataset includes temperature values ​​measured on February 1st regardless of the year, and the 365th dataset includes temperature values ​​measured on December 31st regardless of the year. As another example, if the time unit for classification is "months," then x could be 12.

[0092] Specifically, the control unit 120 can determine the first to the nth groups from the first to the xth datasets, each corresponding to a time period from the first to the nth. The control unit 120 can classify each dataset in the first to the xth datasets that has time information belonging to the jth time period into the jth group. For example, if m=365 and the first time period = January 10 to February 9, then the first group can include all temperature values ​​from the 10th to the 40th datasets.

[0093] Furthermore, the control unit 120 can set a representative value of the temperature values ​​included in the j-th group as the j-th reference temperature value. Here, the representative value can be the mean, median, mode, or a value based on a user-defined rule. Preferably, the representative temperature value of the actual observed histogram and the first to nth reference temperature values ​​can be set according to the same standard.

[0094] The control unit 120 can determine the first to nth temperature correction values ​​based on the representative temperature value and the first to nth reference temperature values.

[0095] Specifically, the control unit 120 can be configured to determine a j-th temperature correction value, which is the difference between the representative temperature value and the j-th reference temperature value among the first to n-th reference temperature values. That is, specifically, the control unit 120 can compare the representative temperature value and the j-th reference temperature value to calculate the difference between the two values, and determine the calculated difference as the j-th temperature correction value.

[0096] The control unit 120 can be configured to correct each temperature value of the actual observed histogram based on the j-th temperature correction value to generate the j-th histogram among the first to nth histograms.

[0097] Specifically, the control unit 120 can compare the representative temperature value with the j-th reference temperature value and determine the method of correction operation based on the comparison result.

[0098] For example, if the j-th reference temperature value exceeds the representative temperature value, the control unit 120 can perform the operation of adding the j-th temperature correction value to each temperature value of the actual observed histogram. Then, the control unit 120 can generate the j-th histogram composed of the corrected temperature values. As another example, if the j-th reference temperature value is less than the representative temperature value, the control unit 120 can perform the operation of subtracting the j-th temperature correction value from each temperature value of the actual observed histogram. Then, the control unit 120 can generate the j-th histogram composed of the corrected temperature values. As another example, if the j-th reference temperature value is equal to the representative temperature value, the actual observed histogram can be set as the j-th histogram as is. In this case, no correction operation needs to be performed.

[0099] The degradation prediction model can be trained using multiple learning datasets collected from multiple other batteries over multiple reference periods.

[0100] Each of the multiple learning datasets may include a reference histogram representing the correspondence between voltage, current, temperature, and frequency values ​​of other batteries during a reference period prior to the actual observation period, and the SOH of other batteries at the beginning of the reference period as features, and may include the SOH of other batteries at the end of the reference period as labels.

[0101] Here, the reference period refers to a predetermined period set for collecting data to train the degradation prediction model, and is the interval before the actual observation period.

[0102] Preferably, the duration of each of the multiple reference time periods can be equal to the total duration of the first to nth time periods. By setting the time scales of the learning dataset and the actual observation dataset to be similar, the degraded prediction model can maintain consistency between the learned data and the data used for prediction, thereby improving the accuracy of the prediction.

[0103] Meanwhile, the batteries used to generate the learning dataset and the batteries used to generate the reference temperature dataset can be the same or different.

[0104] The reference time period and the learning dataset can correspond to each of the multiple batteries. For example, a first to k reference time period can be set corresponding to the first to k batteries (i.e., multiple other batteries), and a first to k learning dataset can be constructed based on the first to k reference time periods. k is a natural number greater than or equal to 2.

[0105] Specifically, the data collected for the m-th cell (which is one of the first to k-th cells) during the m-th reference time period can constitute the m-th learning dataset. The m-th learning dataset may include a m-th reference histogram representing the correspondence between voltage, current, temperature, and frequency values ​​during the m-th reference time period, the state of harmonics (SOH) of the m-th cell at the beginning of the m-th reference time period, and the state of harmonics (SOH) of the m-th cell at the end of the m-th reference time period.

[0106] More specifically, the m-th reference histogram and the SOH of the m-th battery at the beginning of the m-th reference time period can be input as features into the degradation prediction model, and the SOH of the m-th battery at the end of the m-th reference time period can be input as a label into the degradation prediction model.

[0107] Figure 4 and Figure 5 This is a diagram schematically illustrating the structure of a degradation prediction model.

[0108] Reference Figure 4 The input data for the degradation prediction model are the first to nth histograms and the actual observed SOH, and the output result is the target SOH.

[0109] The degradation prediction model can be configured as a CNN-based model. For example, the degradation prediction model can be configured as a 3D-CNN-based deep learning model.

[0110] The degradation prediction model can be configured to receive the first to nth histograms and the actual observed SOH as input data and predict the target SOH.

[0111] Reference Figure 5 This can illustrate the data processing procedure of the degradation prediction model using 3D-CNN.

[0112] The degradation prediction model can include an input layer, convolutional layers, fully connected layers, and an output layer. The input layer takes the first through nth histograms as input data and sends them to the convolutional layers. Each convolutional layer extracts a feature map by performing a convolution operation on the data input to it and reduces the size of the feature map through pooling. The feature map extracted from the last convolutional layer can be converted into a 1D vector and connected to the fully connected layer. The fully connected layer can perform various functions such as classification or regression, similar to a general neural network model. The output layer can use functions such as the softmax function to output the prediction result (i.e., the target SOH).

[0113] Degradation prediction models can progressively reduce the data and extract feature maps from the input data by repeatedly performing 3D convolution operations, batch normalization, ReLU (Revised Linear Unit) activation function operations, and / or max pooling operations on the first to nth histograms of the input. Feature maps extracted from the convolutional and pooling layers of the CNN can be connected and integrated through fully connected layers.

[0114] The feature data generated in the fully connected layer can be combined with the actual observed SOH. Here, combination means that the actual observed SOH is included as an additional input value to the nodes of the fully connected layer, thereby increasing the number of input nodes of the fully connected layer.

[0115] The data with actual observed SOH is processed through fully connected layers, batch normalization, and / or ReLU activation functions, and the target SOH is finally output. The principles and operations of 3D-CNN are widely known, so further explanation will be omitted.

[0116] SOH prediction device 100 utilizes a 3D-CNN-based degradation prediction model to more accurately and reliably predict the SOH of the battery. Specifically, 3D-CNN can effectively learn the complex nonlinear relationships between voltage, current, and temperature within the input data and automatically extract feature maps. This allows for precise analysis of battery state changes, thereby achieving more accurate and reliable prediction of the target SOH.

[0117] The control unit 120 can be configured to predict the degradation rate of the battery in the first to nth time periods based on the actual observed SOH and the target SOH.

[0118] Here, the degradation rate refers to the rate or speed at which the state of oxygen (SOH) of the battery decreases during the first to nth time periods.

[0119] Specifically, the control unit 120 can calculate the difference between the actual observed SOH and the target SOH. Furthermore, the control unit 120 can calculate the degradation rate by dividing the calculated SOH difference by the total time length from the first to the nth time period. The degradation rate numerically represents the trend of SOH decline in the battery over future usage periods.

[0120] The predicted degradation rate can be used for long-term battery performance evaluation and management. For example, the control unit 120 can compare the degradation rate with a preset threshold and set battery usage conditions based on the comparison result. Specifically, if the degradation rate exceeds the threshold, the control unit 120 can adjust the charging / discharging mode or change the usage conditions. As another example, the control unit 120 can monitor changes in the degradation rate and set battery usage conditions if an abnormal change in the degradation rate is predicted.

[0121] The SOH prediction device 100 can improve the safety and reliability of batteries by predicting performance degradation during long-term use based on the degradation rate and setting optimized usage conditions for the battery accordingly.

[0122] The SOH prediction device 100 according to this disclosure can be applied to a BMS. That is, a BMS according to this disclosure may include the aforementioned SOH prediction device 100. In this configuration, at least some components of the SOH prediction device 100 can be implemented by supplementing or adding the functionality of components included in a conventional BMS. For example, the data acquisition unit 110 and the control unit 120 of the SOH prediction device 100 can be implemented as components of the BMS.

[0123] Figure 6 This is a schematic diagram illustrating a battery pack 10 according to another embodiment of the present disclosure.

[0124] The SOH prediction device 100 according to this disclosure can be disposed in the battery pack 10. That is, the battery pack 10 according to this disclosure may include the above-mentioned SOH prediction device 100 and at least one battery cell. In addition, the battery pack 10 may also include electrical components (relays, fuses, etc.) and a housing.

[0125] The positive terminal of battery 11 can be connected to the positive terminal P+ of battery pack 10, and the negative terminal of battery 11 can be connected to the negative terminal P- of battery pack 10.

[0126] The measuring unit 12 can be connected to a first sensing line SL1, a second sensing line SL2, and a third sensing line SL3. Specifically, the measuring unit 12 can be connected to the positive terminal of the battery 11 via the first sensing line SL1 and to the negative terminal of the battery 11 via the second sensing line SL2. The measuring unit 12 can measure the voltage of the battery 11 based on the voltage measured at each of the first sensing line SL1 and the second sensing line SL2.

[0127] Furthermore, the measurement unit 12 can be connected to the current measurement unit A via the third sensing line SL3. For example, the current measurement unit A can be an ammeter or a shunt resistor capable of measuring the charging and discharging currents of the battery 11. The measurement unit 12 can calculate the charge amount by measuring the charging current of the battery 11 via the third sensing line SL3. Furthermore, the measurement unit 12 can calculate the discharge amount by measuring the discharging current of the battery 11 via the third sensing line SL3.

[0128] The data acquisition unit 110 can be connected to the measurement unit 12 via a wired and / or wireless connection to enable communication. The data acquisition unit 110 can receive voltage, current, and / or temperature information of the battery 11 from the measurement unit 12.

[0129] Figure 7 This is a schematic diagram illustrating a vehicle 1 according to yet another embodiment of the present disclosure.

[0130] Reference Figure 7 The above reference Figure 6 The battery pack 10 described can be included in a vehicle 1, such as an electric vehicle (EV) or a hybrid vehicle (HV). Furthermore, the battery pack 10 can drive the vehicle 1 by supplying power to a motor via an inverter included in the vehicle 1. Here, the battery pack 10 may include a State of Health (SOH) prediction device 100. In this case, the SOH prediction device 100 may be an on-board device included in the vehicle 1.

[0131] Figure 8 This is a schematic diagram illustrating server 2 according to yet another embodiment of the present disclosure.

[0132] Reference Figure 8 The SOH prediction device 100 according to this disclosure can be equipped on a server 2. The server 2 provides high-performance computing resources and data storage capabilities to predict the target SOH of the battery.

[0133] The SOH prediction device 100, mounted on server 2, can perform integrated management of a battery system comprising multiple batteries by linking with multiple BMS 3, user terminals 4, and / or vehicle control systems 5. Server 2 can communicate with multiple BMS 3 and / or user terminals 4 via wired and / or wireless connections.

[0134] Server 2 can link with BMS 3 to send SOH prediction results, battery degradation rate, and / or diagnostic results to the corresponding BMS 3 in real time. Alternatively, if the battery condition is diagnosed as abnormal, Server 2 can send a warning or control signal to the corresponding BMS 3.

[0135] Server 2 can be linked to user terminal 4 so that the user can remotely monitor the battery status. The user can use a dedicated application to check the battery status in real time.

[0136] Figure 9 This is a diagram schematically illustrating a SOH prediction method according to yet another embodiment of the present disclosure. Figure 10 It is shown schematically. Figure 9 The diagram shows the sub-steps of step S920.

[0137] Each step of the SOH prediction method can be performed by the SOH prediction device 100. In the following text, for ease of explanation, content overlapping with the above will be omitted or briefly described.

[0138] Reference Figures 1 to 9 The SOH prediction method includes steps S910, S920 and S930.

[0139] Step S910 is to obtain an actual observation histogram representing the correspondence between the voltage, current and temperature values ​​of the battery during the actual observation period, and can be performed by the data acquisition unit 110.

[0140] Step S920 is a step of generating the first to nth histograms corresponding to the first to nth time periods after the actual observation period based on the actual observation histogram and the reference temperature dataset, and can be executed by the control unit 120.

[0141] Reference Figure 10 Step S920 may include steps S1010, S1020, S1030 and S1040.

[0142] Step S1010 is the step of generating a reference temperature dataset based on multiple temperature datasets, and can be performed by the control unit 120. For example, the control unit 120 can be configured to generate the reference temperature dataset based on multiple temperature datasets collected from multiple other batteries during past data collection periods with a time length greater than or equal to the total time length of the first to nth time periods.

[0143] Step S1020 is a step of comparing the representative temperature value of the actual observed histogram with each of the first to nth reference temperature values, and can be performed by the control unit 120.

[0144] Step S1030 is a step of calculating the difference between the representative temperature value of the actual observed histogram and the first to nth reference temperature values ​​to determine the first to nth temperature correction values, and can be executed by the control unit 120.

[0145] Step S1040 is the step of generating the first to nth histograms by correcting the temperature values ​​of the actual observed histograms based on the first to nth temperature correction values, and can be executed by the control unit 120.

[0146] Specifically, the control unit 120 can compare the representative temperature value with the j-th reference temperature value and determine the method of correction operation based on the comparison result.

[0147] For example, if the j-th reference temperature value exceeds the representative temperature value, the control unit 120 can perform the operation of adding the j-th temperature correction value to each temperature value in the actual observation histogram. Then, the control unit 120 can generate the j-th histogram composed of the corrected temperature values. As another example, if the j-th reference temperature value is less than the representative temperature value, the control unit 120 can perform the operation of subtracting the j-th temperature correction value from each temperature value in the actual observation histogram. Then, the control unit 120 can generate the j-th histogram composed of the corrected temperature values. As yet another example, if the j-th reference temperature value is equal to the representative temperature value, the actual observation histogram can be set as the j-th histogram as is. In this case, no correction operation needs to be performed.

[0148] Step S930 is a step of predicting the target SOH of the battery at the end of the nth time period based on a pre-set degradation prediction model, according to the first to nth histograms and the actual observed SOH representing the SOH of the battery at the end of the actual observation period, and can be executed by the control unit 120.

[0149] Specifically, the control unit 120 can predict the target SOH by inputting the first to nth histograms and the actual observed SOH into the degradation prediction model.

[0150] The degradation prediction model can be a deep learning model trained by using a histogram representing the correspondence between voltage, current, temperature and frequency values, as well as the battery's state of harm (SOH) at a specific time point, as input to the learning dataset.

[0151] For example, a degradation prediction model can be a model trained by taking a reference histogram representing the correspondence between voltage, current, temperature and frequency values ​​during a reference period and the battery's state of harm (SOH) at the end of the reference period as input to a learning dataset.

[0152] The control unit 120 can be configured to change the preset usage conditions for the battery based on the predicted target SOH or degradation rate.

[0153] Specifically, the control unit 120 can appropriately change the preset usage conditions to correspond to the state of the battery.

[0154] Here, the usage conditions may include at least one of the maximum permissible temperature, upper limit of SOC, lower limit of SOC, and upper limit of C rate.

[0155] In one embodiment, the control unit 120 can compare the target SOH with a preset threshold SOH and change the usage conditions based on the comparison result.

[0156] Here, the threshold SOH can be defined as the SOH used as a standard to determine the time point at which the likelihood of battery performance degradation or safety degradation increases. The threshold SOH can be preset theoretically and / or experimentally.

[0157] For example, if the target SOH is lower than or equal to the threshold SOH, the control unit 120 may perform control to reduce at least one of the maximum permissible temperature, the upper limit of SOC, and the upper limit of C rate. As another example, if the target SOH is lower than or equal to the threshold SOH, the control unit 120 may perform control to increase the lower limit of SOC.

[0158] Furthermore, the degree of change in usage conditions can be adjusted to be proportional to the degree of battery degradation. Specifically, the level of adjustment of the applied usage conditions can be determined according to a control protocol that establishes the correspondence between the State of Health (SOH) indicator and the degree of adjustment of the usage conditions.

[0159] The control protocol includes adjustment coefficients or adjustment plots set for each SOH range, and can be configured such that, for example, when the target SOH is 90% or higher, normal conditions are maintained without adjustment; when the target SOH is 80% or higher but less than 90%, the C rate is reduced by 10%; and when the target SOH is less than 80%, the C rate is reduced by 20%. The control protocol can be predefined based on experimental data, simulation results, or system-specific requirements.

[0160] In another embodiment, the degradation rate can be compared with a preset threshold rate, and the usage conditions can be changed based on the comparison result.

[0161] Here, the threshold rate can be defined as the state of harmonics (SOH) used as a standard to determine the point at which the likelihood of accelerated battery degradation increases. Accelerated degradation can refer to a state in which the rate of SOH decline increases rapidly, potentially leading to shortened battery life or safety issues. The threshold rate can be preset theoretically and / or experimentally.

[0162] For example, if the degradation rate is greater than or equal to the threshold rate, the control unit 120 may perform control to reduce at least one of the maximum permissible temperature, the upper limit of SOC, and the upper limit of C rate. As another example, if the degradation rate is greater than or equal to the threshold rate, the control unit 120 may perform control to increase the lower limit of SOC.

[0163] In another embodiment, the control unit 120 can be configured to output an alarm when the target SOH is lower than or equal to a threshold SOH or when the degradation rate is higher than or equal to a threshold rate. That is, the control unit 120 can immediately output an alarm to notify external parties of the battery status. For example, the control unit 120 can output an alarm to notify battery status to an alarm unit (not shown), a display unit (not shown), a user terminal (not shown), and a server that are connected via wired and / or wireless means to enable communication.

[0164] According to embodiments of this disclosure, the SOH prediction device 100 can increase the expected lifespan of the battery and prevent safety accidents caused by abnormal battery degradation in advance by taking appropriate measures based on the prediction results.

[0165] Another embodiment of this disclosure may provide a computer-readable storage medium having programs recorded thereon for executing the various embodiments described above on a computer.

[0166] A program can be implemented as a hardware component, a software component, and / or a combination of hardware and software components. The program can be executed by any system capable of executing computer-readable instructions.

[0167] Software may include computer programs, code, instructions, or combinations thereof, which may configure processing equipment to perform desired operations or may independently or jointly command processing equipment.

[0168] Software can be implemented as a computer program that includes instructions stored on a computer-readable storage medium. Examples of computer-readable storage media include magnetic storage media (e.g., read-only memory (ROM), random access memory (RAM), floppy disks, hard disks, etc.) and optically readable media (e.g., CD-ROMs, DVDs (Digital Universal Discs), etc.). Computer-readable storage media can be distributed across network-connected computer systems, allowing computer-readable code to be stored and executed in a distributed manner. The storage medium can be read by a computer, stored in memory, and executed by a processor.

[0169] Computer-readable storage media may be provided in the form of non-transitory storage media. Here, the term "non-transitory storage media" simply means that it is a tangible device and does not contain signals (e.g., electromagnetic waves), and the term does not distinguish between cases where data is stored semi-permanently on the storage medium and cases where data is temporarily stored. For example, a "non-transitory storage medium" may include a buffer in which data is temporarily stored.

[0170] Furthermore, programs can be provided as part of a computer program product. Computer program products can be traded as commodities between sellers and buyers.

[0171] Computer program products may include software programs and computer-readable storage media storing the software programs. For example, a computer program product may include a product in the form of a software program (e.g., a downloadable application) distributed electronically by a manufacturer of electronic devices or through an electronic marketplace. For electronic distribution, at least a portion of the software program may be stored on a storage medium or temporarily generated. In this case, the storage medium may be the storage medium of a server belonging to the manufacturer of the electronic devices, a server of an electronic marketplace, or a relay server temporarily storing the software program.

[0172] The embodiments of this disclosure described above can be implemented not only by means of apparatus and methods, but also by a program that implements functions corresponding to the configuration of the embodiments of this disclosure, or a recording medium on which the program is recorded. The program or recording medium can be readily implemented by those skilled in the art from the above description of the embodiments.

[0173] This disclosure has been described in detail. However, it should be understood that while the detailed description and specific examples indicate preferred embodiments of this disclosure, they are given by way of illustration only, as various changes and modifications within the scope of this disclosure will become apparent to those skilled in the art from the detailed description.

[0174] Furthermore, without departing from the technical aspects of this disclosure, those skilled in the art can make many substitutions, modifications and changes to the disclosure described above, and this disclosure is not limited to the above embodiments and drawings, and each embodiment can be selectively combined in part or in whole to allow for various modifications.

[0175] (Explanation of the labels in the attached diagram)

[0176] 1: Vehicle

[0177] 2: Server

[0178] 3: BMS

[0179] 4: User terminal

[0180] 10: Battery Pack

[0181] 11: Battery

[0182] 12: Measurement Unit

[0183] 100: SOH Prediction Device

[0184] 110: Data Acquisition Unit

[0185] 120: Control Unit

[0186] 130: Storage unit

Claims

1. A SOH prediction device, comprising: A data acquisition unit is configured to acquire an actual observation histogram representing the correspondence between voltage, current, temperature and frequency values ​​of the battery during the actual observation period. as well as A control unit is configured to generate first to nth histograms corresponding to the first to nth time periods following the actual observation period based on the actual observation histograms and a reference temperature dataset, and to predict a target SOH representing the SOH of the battery at the end of the nth time period based on a pre-set degradation prediction model, according to the first to nth histograms and the actual observed SOH representing the SOH of the battery at the end of the actual observation period, where n is a natural number greater than or equal to 2.

2. The SOH prediction device according to claim 1, in, The reference temperature dataset includes the first to nth reference temperature values ​​corresponding to the first to nth time periods, and The control unit is configured to generate the first to nth histograms based on a comparison between the representative temperature value of the actual observed histogram and each of the first to nth reference temperature values.

3. The SOH prediction device according to claim 2, in, The control unit is configured to generate the reference temperature dataset based on multiple temperature datasets collected from multiple other batteries during past data collection periods with a time length greater than or equal to the total time length of the first to nth time periods.

4. The SOH prediction device according to claim 2, in, The control unit is configured to: Determine the j-th temperature correction value, wherein the j-th temperature correction value is the difference between the representative temperature value and the j-th reference temperature among the first to nth reference temperature values, and Each temperature value in the actual observed histogram is corrected based on the j-th temperature correction value to generate the j-th histogram in the first to n-th histograms, where j is a natural number less than or equal to n.

5. The SOH prediction device according to claim 1, in, The degradation prediction model is trained using multiple learning datasets collected from multiple other batteries over multiple reference periods, and Each of the plurality of learning datasets includes a reference histogram representing the correspondence between voltage, current, temperature and frequency values ​​of the other batteries during the reference period prior to the actual observation period, and the SOH of the other batteries at the beginning of the reference period as features, and includes the SOH of the other batteries at the end of the reference period as labels.

6. The SOH prediction device according to claim 5, in, The duration of each of the plurality of reference time periods is equal to the total duration of the first to nth time periods.

7. The SOH prediction device according to claim 1, in, The duration of each of the first to nth time periods is equal to the duration of the actual observation period.

8. The SOH prediction device according to claim 1, in, The degradation prediction model is configured as a model based on a convolutional neural network (CNN).

9. The SOH prediction device according to claim 1, in, The control unit is configured to predict the degradation rate of the battery during the first to nth time periods based on the actual observed SOH and the target SOH.

10. A battery pack comprising an SOH prediction device according to any one of claims 1 to 9.

11. A vehicle comprising a SOH prediction device according to any one of claims 1 to 9.

12. A server comprising an SOH prediction device according to any one of claims 1 to 9.

13. A method for predicting SOH, comprising: Obtain the actual observation histogram representing the correspondence between the voltage, current, temperature, and frequency values ​​of the battery during the actual observation period; Based on the actual observation histogram and the reference temperature dataset, generate the first to nth histograms corresponding to the first to nth time periods after the actual observation period; as well as Based on a pre-set degradation prediction model, and according to the first to nth histograms and the actual observed SOH representing the SOH of the battery at the end of the actual observation period, a target SOH representing the SOH of the battery at the end of the nth period is predicted, where n is a natural number greater than or equal to 2.

Images