Machine learning-based battery charge / discharge current prediction method and device for charge / discharge strategy for battery thermal management

Abstract

The present invention provides a method by which a computing device predicts a battery charge / discharge current, comprising the steps of: acquiring, according to a charge / discharge current profile of a battery, training data and verification data including a temperature profile; using the training data to train a machine learning model; using the verification data to verify and retrain the trained machine learning model; and using the machine learning model to predict a current profile according to the temperature profile of the battery.

Description

A machine learning-based battery charge / discharge current prediction method and device for battery thermal management.

[0001] The present invention relates to a method and device for predicting battery charge and discharge current based on machine learning for a charge and discharge strategy for battery thermal management.

[0002] Secondary batteries, unlike non-rechargeable primary batteries, are rechargeable and dischargeable. Low-capacity secondary batteries are used in small, portable electronic devices such as smartphones, feature phones, laptops, digital cameras, and camcorders, while large-capacity secondary batteries are widely used as power sources for motor drives and power storage in hybrid and electric vehicles. These secondary batteries include an electrode assembly comprising a positive and negative electrode, a case housing the electrode assembly, and electrode terminals connected to the electrode assembly.

[0003] Although there is a lot of research being conducted on forward prediction technology, such as predicting the State of Charge (SoC) of a battery using parameters such as current, voltage, or temperature of a battery including such secondary cells, there is insufficient research on reverse prediction technology for the next operation strategy based on the current temperature of the battery in terms of the battery charge / discharge strategy.

[0004] The above-described information disclosed in the background technology of this invention is only intended to improve understanding of the background of the present invention, and therefore may include information that does not constitute prior art.

[0005] The problem to be solved by the present invention is to provide a method and device for predicting battery charge and discharge current based on machine learning of a charge and discharge strategy for battery thermal management.

[0006] However, the technical problems to be solved by the present invention are not limited to the problems described above, and other problems not mentioned can be clearly understood by those skilled in the art from the description of the invention described below.

[0007] In order to solve the above technical problem, a battery charge / discharge current prediction method performed by a computing device according to an embodiment of the present invention is provided, comprising the steps of: acquiring learning data and verification data including a temperature profile according to a charge / discharge current profile of a battery; learning a machine learning model using the learning data; verifying and relearning the learned machine learning model using the verification data; and predicting a current profile according to a temperature profile of a battery using the machine learning model.

[0008] In one example, the step of predicting a current profile according to a temperature profile of the battery may include a step of predicting a change in charge / discharge current of the battery over time that satisfies a change in temperature of the battery over time.

[0009] According to another example, the step of acquiring the learning data and verification data may include a step of acquiring a plurality of data on changes in the charge / discharge current of the battery over time and changes in the temperature of the battery according to changes in the charge / discharge current of the battery.

[0010] According to another example, the step of obtaining the learning data and verification data may include a step of obtaining a plurality of data by random sampling that extracts variable combinations having a homogeneous distribution within a given variable combination and range.

[0011] According to another example, the step of obtaining the training data and the verification data may include the step of obtaining the training data and the verification data, each having the same distribution but including different current and temperature data.

[0012] In another example, the step of training the machine learning model includes: the temperature of the battery over time [T(t)], the first order change in the temperature of the battery [ T(t)], the second order change in the temperature of the battery[ The method may include a step of learning the machine learning model that receives the state of charge of the battery over time [SoC(t)] and the state of charge of the battery over time as input values ​​and outputs the current of the battery over time as an output value.

[0013] According to another example, the first order change in temperature of the battery [ T(t)] and the second order change in temperature of the battery [ [T(t)] can have the following relationship with the temperature of the battery over time [T(t)].

[0014] 1st order change:

[0015] 2nd order change:

[0016] According to another example, the state of charge of the battery over time [SOC(t)] is the predicted current of the battery over time [ ] can be expressed as the following formula.

[0017]

[0018] According to another example, the step of verifying and retraining the machine learning model may include the step of predicting a predicted value for the current of the battery based on the temperature of the battery through the machine learning model using the verification data that is not included in the training data, the step of comparing the predicted value for the current with the current data of the verification data, and the step of comparing the temperature data of the battery obtained from the predicted value for the current with the temperature data of the battery of the verification data.

[0019] As a technical means for achieving the above-described technical tasks, a computer program stored in a recording medium is provided to execute the above-described method using a computing device.

[0020] In order to solve the above technical problem, according to one embodiment of the present invention, a battery charge / discharge current prediction device is provided, comprising: a memory for storing learning data and verification data including a temperature profile according to a charge / discharge current profile of a battery; and at least one processor for acquiring the learning data and the verification data, learning a machine learning model using the learning data, verifying and relearning the learned machine learning model using the verification data, and predicting a current profile according to a temperature profile of the battery using the machine learning model.

[0021] For example, the processor can predict the change in charge and discharge current of the battery over time that satisfies the change in temperature of the battery over time.

[0022] According to another example, the processor can obtain a plurality of data about changes in the charge / discharge current of the battery over time and changes in the temperature of the battery according to changes in the charge / discharge current of the battery.

[0023] According to another example, the processor can calculate the temperature of the battery over time [T(t)], the first order change in the temperature of the battery [ T(t)], the second order change in the temperature of the battery[ The machine learning model can be trained by inputting the state of charge of the battery [SoC(t)] over time and the current of the battery over time as an output value.

[0024] According to another example, the processor may predict a predicted value for current of the battery based on the temperature of the battery through the machine learning model using the verification data that is not included in the training data, compare the predicted value for current with the current data of the verification data, and compare the temperature data of the battery obtained from the predicted value for current with the temperature data of the battery of the verification data.

[0025] According to the present invention, a machine learning-based reverse prediction model that predicts a charge / discharge profile satisfying a given battery temperature profile can be secured, and in the process, the amount of learning data can be sequentially increased to enhance learning efficiency.

[0026] However, the effects that can be obtained through the present invention are not limited to the effects described above, and other technical effects that are not mentioned can be clearly understood by those skilled in the art from the description of the invention described below.

[0027] The following drawings attached to this specification illustrate preferred embodiments of the present invention, and together with the detailed description of the invention described below, serve to further understand the technical idea of ​​the present invention, and therefore, the present invention should not be interpreted as being limited to matters described in such drawings.

[0028] FIG. 1 is a drawing schematically illustrating a battery pack according to one embodiment of the present invention.

[0029] FIG. 2 is a flowchart for explaining a battery charge / discharge current prediction method according to one embodiment of the present invention.

[0030] FIG. 3 is a drawing for explaining a charge state profile and a current profile corresponding to the charge state profile according to one embodiment of the present invention.

[0031] FIG. 4 and FIG. 5 are drawings for explaining a battery charge / discharge current prediction method according to one embodiment of the present invention.

[0032] Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the attached drawings. Prior to this, terms or words used in this specification and claims should not be interpreted as limited to their typical or dictionary meanings, but should be interpreted with meanings and concepts that conform to the technical idea of ​​the present invention based on the principle that the inventor can appropriately define the concept of the term in order to explain his own invention in the best way. Therefore, it should be understood that the embodiments described in this specification and the configurations illustrated in the drawings are only some of the most preferred embodiments of the present invention and do not represent all of the technical idea of ​​the present invention, and various equivalents and modifications may exist at the time of filing this application. In addition, when used in this specification, "comprise" and "include" and / or "comprising" specify the presence of mentioned shapes, numbers, steps, operations, elements, components and / or groups thereof, and do not exclude the presence or addition of one or more other shapes, numbers, operations, elements, components and / or groups. Additionally, when describing embodiments of the present invention, “may” and “may be” may include “one or more embodiments of the present invention.”

[0033] Additionally, to facilitate understanding of the invention, the attached drawings are not drawn to scale and some components may be exaggerated in size. Furthermore, identical components may be assigned the same reference numbers in different embodiments.

[0034] The statement that two compared objects are "identical" means "substantially identical." Therefore, "substantially identical" may include deviations considered low in the art, such as deviations of less than 5%. Furthermore, uniformity of a parameter over a given region may imply uniformity on average.

[0035] Although terms like "first" and "second" are used to describe various components, these components are not limited by these terms. These terms are used merely to distinguish one component from another, and unless otherwise specified, a "first" component may also be a "second" component.

[0036] Throughout the specification, unless otherwise specifically stated, each element may be singular or plural.

[0037] Any configuration being placed “on top (or bottom)” of a component or “on top (or bottom)” of a component may mean not only that any configuration is placed in contact with the top (or bottom) surface of said component, but also that other configurations may intervene between said component and any configuration placed on (or under) said component.

[0038] Additionally, when it is described that a component is “connected,” “coupled,” or “connected” to another component, it should be understood that the components may be directly connected or connected to each other, but that other components may also be “interposed” between the components, or that each component may be “connected,” “coupled,” or “connected” through another component. Furthermore, when it is said that a part is electrically coupled to another part, this includes not only cases where they are directly connected, but also cases where they are connected with another element in between.

[0039] When reference is made throughout the specification to “A and / or B,” this means A, B, or A and B, unless otherwise stated. In other words, “and / or” includes all or any combination of the listed items. When reference is made to “C through D,” this means C or more and D or less, unless otherwise stated.

[0040] FIG. 1 schematically illustrates a battery pack according to one embodiment of the present invention.

[0041] Referring to FIG. 1, a battery pack (100) includes a battery module (110), a processor (150), a memory (160), a voltage measuring unit (120), a current measuring unit (130), and a temperature measuring unit (140).

[0042] A battery pack (100) includes at least one battery module (110) and a pack housing having a receiving space for accommodating at least one battery module (110). The battery module (110) may include a plurality of battery cells (111) and a module housing. The battery cells (111) may be accommodated inside the module housing in a stacked form. The battery cells (111) may be provided with a positive lead and a negative lead. Depending on the battery shape, a circular type, a square type, or a pouch type battery cell may be used as the battery cell.

[0043] The above battery pack (100) may be configured such that a single stack of stacked cells constitutes a module instead of the battery module. The cell stack may be accommodated in a receiving space of the pack housing or in a receiving space partitioned by a frame, bulkhead, or the like.

[0044] The above battery cell (111) generates a large amount of heat during charging / discharging. The generated heat accumulates in the battery cell (111) and accelerates the deterioration of the battery cell (111). Therefore, the battery pack (100) further includes a cooling member to suppress the deterioration of the battery cell. The cooling member is provided at the bottom of the receiving space where the battery cell (111) is provided, but is not limited thereto, and may also be provided at the top or side depending on the battery pack (100).

[0045] The above battery cell (111) may be subject to exhaust gas discharged from the inside of the battery cell to the outside of the battery cell in the event of an abnormal operating condition, also known as thermal runaway or thermal event. The battery pack (100) or the battery module (110) may be provided with an exhaust port or the like for exhaust gas discharge to prevent damage to the battery pack or module.

[0046] The battery module (110) includes at least one battery cell (111), and the battery cell (111) may be a rechargeable secondary battery. For example, the battery cell (111) may include at least one selected from the group consisting of a nickel-cadmium battery, a lead-acid battery, a nickel metal hydride (NiMH) battery, a lithium-ion battery, a lithium polymer battery, and the like.

[0047] The number and connection method of battery cells (111) included in the battery module (110) can be determined based on the amount of power and voltage required for the battery pack (100), etc. Although the battery cells (111) included in the battery module (110) are shown as being connected in series in FIG. 1 for conceptual purposes only, the battery cells (111) may be connected in parallel with each other, or may be connected in series and in parallel. Although the battery pack (100) is shown in FIG. 1 for conceptual purposes only as including one battery module (110), it may include a plurality of battery modules (110) connected in series, parallel, or both in series and in parallel. The battery module (110) may also include only one battery cell (111).

[0048] The battery module (110) may include a plurality of battery modules, each of which is composed of a plurality of battery cells (111). The battery pack (100) includes a pair of pack terminals (101, 102) to which an electrical load or charging device can be connected.

[0049] A battery pack (100) according to one embodiment of the present invention may include a switch. The switch may be connected between a battery module (110) and one of the pack terminals (101, 102) (e.g., 101). The switch may be controlled by a processor (150). Although not shown in FIG. 1, the battery pack (100) may further include a battery protection circuit, a fuse, a current sensor, and the like.

[0050] A battery charge / discharge current prediction device according to one embodiment of the present invention includes a processor (150) and a memory (160).

[0051] The processor (150) controls the overall operation of the battery charge / discharge current prediction device. For example, the processor (150) may be implemented in a form that optionally includes a processor, an application-specific integrated circuit (ASIC), another chipset, a logic circuit, a register, a communication modem, and / or a data processing device known in the art to perform the above-described operation.

[0052] The processor (150) can perform basic arithmetic, logic, and input / output operations, and execute program codes stored in, for example, memory (160). The processor (150) can store data in the memory (160) or load data stored in the memory (160).

[0053] The memory (160) is a recording medium that can be read by the processor (150), and may include a permanent mass storage device such as a RAM, a ROM, and a disk drive. The memory (160) may store an operating system and at least one program or application code. The memory (160) may store a program code for predicting a current profile according to a temperature profile of a battery using a machine learning model according to an embodiment of the present invention. For example, the memory (160) may store learning data and verification data including a temperature profile according to a charge / discharge current profile of the battery. For example, the memory (160) may store data generated by measuring at least one parameter of the battery module (110). In addition, the memory (160) may store a specific capacity of the battery module (110). In addition, the memory (160) may store SOC-OCV data of the battery module (110). For example, the data may include the charge / discharge current, terminal voltage, and / or temperature of the battery. Additionally, the data may include the discharge rate of the battery. The memory (160) may store program code and SOC-OCV data for estimating the SOC of the battery using data generated by measuring at least one parameter of the battery module (110). The at least one parameter of the battery module (110) refers to a component or variable such as the terminal voltage, charge / discharge current, and / or ambient temperature of the battery module (110).

[0054] A battery charge / discharge current prediction device according to one embodiment of the present invention may further include a voltage measurement unit (120), a current measurement unit (130), and a temperature measurement unit (140) for measuring at least one parameter of a battery module (110). A battery charge / discharge current prediction device according to one embodiment of the present invention may further include a communication module for communicating with other devices, such as an electronic control unit of a vehicle, a controller of a charging device, etc.

[0055] The voltage measuring unit (120) may be configured to measure the voltage of the battery module (110). For example, as illustrated in the configuration of FIG. 1, the voltage measuring unit (120) may be electrically connected to both ends of the battery module (110) and / or the battery cell (111). In addition, the voltage measuring unit (120) may be electrically connected to the processor (150) so as to transmit and receive electrical signals. In addition, the voltage measuring unit (120) may measure the voltage across both ends of the battery module (110) and / or the battery cell (111) at time intervals under the control of the processor (150) and output a signal representing the magnitude of the measured voltage to the processor (150). At this time, the processor (150) may determine the voltage of the battery module (110) and / or the battery cell (111) from the signal output from the voltage measuring unit (120). For example, the voltage measurement unit (120) can be implemented using a voltage measurement circuit commonly used in the art.

[0056] In addition, the current measuring unit (130) may be configured to measure the current of the battery. For example, as illustrated in the configuration of FIG. 1, the current measuring unit (130) may be electrically connected to a current sensor provided on the charge / discharge path of the battery module (110) and / or the battery cell (111). In addition, the current measuring unit (130) may be electrically connected to the processor (150) so as to exchange electrical signals. In addition, the current measuring unit (130) may repeatedly measure the magnitude of the charge current or discharge current of the battery module (110) and / or the battery cell (111) at time intervals under the control of the processor (150) and output a signal representing the magnitude of the measured current to the processor (150). At this time, the processor (150) may determine the magnitude of the current from the signal output from the current measuring unit (130). For example, the current sensor may be implemented using a Hall sensor or sense resistor commonly used in the art.

[0057] The temperature measuring unit (140) may be configured to measure the temperature of the battery. For example, as illustrated in the configuration of FIG. 1, the temperature measuring unit (140) may be connected to the battery module (110) and / or the battery cell (111) to measure the temperature of the secondary battery provided in the battery module (110) and / or the battery cell (111). In addition, the temperature measuring unit (140) may be electrically connected to the processor (150) so as to exchange electrical signals. In addition, the temperature measuring unit (140) may repeatedly measure the temperature of the secondary battery at time intervals and output a signal representing the magnitude of the measured temperature to the processor (150). At this time, the processor (150) may determine the temperature of the secondary battery from the signal output from the temperature measuring unit (140). For example, the temperature measuring unit (140) may be implemented using a thermocouple commonly used in the art.

[0058] In addition, the processor (150) may monitor and calculate the state (voltage, current, temperature, state of charge (SOC), state of health (SOH), etc.) of the battery module by using at least one of the voltage measurement value, current measurement value, and temperature measurement value for the battery module (110) received from the voltage measurement unit (120), the current measurement unit (130), and the temperature measurement unit (140). Here, the SOC may be obtained as a numerical value corresponding to the remaining amount of the battery module (110) in the range of 0% to 100%. In addition, the processor (150) may perform a control function (e.g., temperature control, balancing control, charge / discharge control, etc.), a protection function (e.g., over-discharge, over-charge, over-current prevention, short-circuit, fire extinguishing function, etc.), etc., based on the state monitoring result. Additionally, the processor (150) may perform wired or wireless communication functions with external devices of the battery pack (e.g., a higher controller or a vehicle or a charger or PCS, etc.).

[0059] The processor (150) may also control the charging / discharging and protection operations of the battery. To this end, the processor (150) may include a charging / discharging control unit, a balancing control unit, and a protection unit.

[0060] A battery management system including a processor (150) is a system that monitors the battery status and performs diagnosis and control, communication, and protection functions, and may calculate a charge / discharge status, calculate a battery life or state of health (SOH: State Of Health), cut off battery power (relay control) when necessary, perform thermal management (cooling, heating, etc.) control, perform a high-voltage interlock function, and detect or calculate insulation and short-circuit conditions.

[0061] In one aspect of the present invention, the processor (150) can estimate the SOC of the battery module (110) by integrating the charge and discharge current of the battery module (110). Here, when charging or discharging of the battery module (110) begins, the initial value of the state of charge can be determined using the open circuit voltage (OCV) of the battery module (110) measured before charging or discharging begins. To this end, the processor (150) can map the state of charge corresponding to the open circuit voltage of the battery module (110) from the SOC-OCV data based on the SOC-OCV data that defines the state of charge for each open circuit voltage.

[0062] <Mathematical Formula 1>

[0063] SoC = SoC i-1 +dSoC

[0064] Here, SOC is the state of charge of the battery, and SoC i-1 It is the initial state of charge of the battery, and dSOC is the accumulated value of the battery's charge and discharge current.

[0065] <Mathematical Formula 2>

[0066]

[0067] Here, dSOC refers to the accumulated value of the battery's charge / discharge current from the start of charging or discharging, and specific capacity refers to the battery's design capacity.

[0068] In another aspect of the present invention, the processor (150) can estimate the SOC of the battery module (110) using an extended Kalman filter. The extended Kalman filter refers to a mathematical algorithm that adaptively estimates the state of charge of a secondary battery using the voltage, current, and temperature of the secondary battery.

[0069] The SOC of the battery module (110) can also be determined by other known methods capable of estimating the state of charge by selectively utilizing the voltage, current, and temperature of the secondary battery, in addition to the above-described current integration method or extended Kalman filter.

[0070] Fig. 2 is a flowchart illustrating a battery charge / discharge current prediction method according to one embodiment of the present invention. The battery charge / discharge current prediction method according to one embodiment of the present invention can be performed by the processor (150) illustrated in Fig. 1.

[0071] Referring to FIG. 2, a method for predicting a battery charge / discharge current according to the present invention includes a step of obtaining learning data and verification data including a temperature profile according to a charge / discharge current profile of a battery (S110), a step of learning a machine learning model using the learning data (S120), a step of verifying and relearning the learned machine learning model using the verification data (S130), and a step of predicting a current profile according to a temperature profile of a battery using the machine learning model (S140).

[0072] According to one embodiment, the step (S110) of acquiring the learning data and verification data may include a step of acquiring a plurality of data on changes in the charge / discharge current of the battery over time and changes in the temperature of the battery according to changes in the charge / discharge current of the battery.

[0073] In addition, according to one embodiment, the step (S110) of acquiring the learning data and verification data may include a step of acquiring a plurality of data by random sampling that extracts a variable combination having a homogeneous distribution within a given variable combination and range.

[0074] In addition, according to one embodiment, the step (S110) of acquiring the learning data and the verification data may include a step of acquiring the learning data (S111) and the verification data (S112) which have the same distribution but each include different current and temperature data.

[0075] A processor (150) according to one embodiment of the present invention can obtain a temperature profile [T(t)] according to various charge / discharge profiles [I(t)] of a battery. Here, t is a time index and can represent information on a point in time according to a certain interval. For example, in an embodiment of the present invention, a profile (t=0, 1,…,72) with an interval of 50 seconds and a total of 3600 seconds can be used.

[0076] For example, a method of acquiring learning data (S111) and verification data (S112) according to one embodiment of the present invention may be a method used to secure data to be used for learning and verifying a machine learning model. Here, the battery temperature for an arbitrary current profile can be secured in various ways, but since a large number of samples are required due to the nature of the machine learning model, data can be collected using COMSOL simulation in the present invention. However, the present invention is not limited thereto, and the present invention does not limit a specific method for acquiring a temperature profile [T(t)] according to a given current profile [I(t)].

[0077] The machine learning-based model construction method proposed in this invention is expected to work as long as there is a consistent correlation between the inputs (e.g., temperature) and outputs (e.g., current) used for learning, regardless of the method used to acquire the training data. However, to meet the required number of training data sets for conventional machine learning methods, it is advantageous to apply a numerical data acquisition technique through thermal fluid analysis.

[0078] In embodiments of the present invention, battery operation was simulated using COMSOL, a commercial thermal fluid analysis program, and current-temperature data was obtained and verified through this simulation. This COMSOL battery model is a computational model verified through experimental measurements. For example, the temperature of the battery verified through experiments is the temperature of the outer wall of the battery can, while the temperature at which the embodiment of the present invention operates may represent the temperature at the exact center of the battery's internal structure.

[0079] Additionally, since it is important to train the model across a wide range of charge / discharge profiles, the training data (S111) and validation data (S112) can be acquired using the Latin-Hypercube sampling method to avoid sample bias. For example, the number of validation data is typically selected to be 1 to 5% of the number of training data.

[0080] For example, a current profile [I(t)] can be extracted using Latin-Hypercube sampling. For example, Latin-Hypercube (LH) sampling can represent a sampling technique that obtains an arbitrary combination of numerical variables with the least bias within a given range. For example, the processor (150) can set the shape of the current profile to a step-function shape, and extract variables that simulate the step-function through the LH sampling technique to ensure that there is no bias in the current range.

[0081] In addition, the processor (150) is based on the idea that when a state of charge [SoC(t)] profile that changes over time is determined, a current profile is determined accordingly, and four variables that simulate the state of charge profile: initial state of charge (SoC0), state of charge at the time of change (SoC c ), final state of charge (SoC1), change point (t c) can be extracted through LH sampling and converted into a current profile for use. For example, each variable can have a range as in mathematical expression 3.

[0082] <Mathematical Formula 3>

[0083] 0 SoC0,SoC t ,SoC1 1

[0084] 300 seconds t c 3300 seconds

[0085] Here, the change time of the current step-function (t c ) is set between 300 and 3300 seconds to prevent rapid changes in the state of charge, i.e., excessive increase in current intensity.

[0086] Referring to Fig. 3, Fig. 3 is a diagram illustrating a charge state profile and a current profile according to the time, and a current profile according to a combination of variables can be prepared according to the corresponding relationship. For example, the current is 0 ~ t c Up to the second, in the century of I0', t c It can flow at the intensity of I1' up to ~3600 seconds.

[0087] According to one embodiment, the step (S120) of learning the machine learning model includes: the temperature of the battery over time [T(t)], the first order change in the temperature of the battery [ T(t)], the second order change in the temperature of the battery[ The method may include a step of learning the machine learning model that receives the state of charge of the battery over time [SoC(t)] and the state of charge of the battery over time as input values ​​and outputs the current of the battery over time as an output value.

[0088] Additionally, according to one embodiment, the first order change in temperature of the battery [ T(t)] and the second order change in temperature of the battery [ [T(t)] can have a relationship with the temperature of the battery over time [T(t)] as in Equation 4.

[0089] <Mathematical Formula 4>

[0090] 1st order change:

[0091] 2nd order change:

[0092] Additionally, according to one embodiment, the state of charge of the battery over time [SoC(t)] is the current of the battery predicted over time [ ] can be expressed as in mathematical formula 5.

[0093] <Mathematical Formula 5>

[0094]

[0095] According to one embodiment of the present invention, a processor (150) can train an LSTM-based machine learning model using training data (S111). For example, the embodiment of the present invention can be implemented using MATLAB's Deep Learning Toolbox. Furthermore, the trained machine learning model (S125) can be stored and utilized in subsequent verification procedures.

[0096] For example, a machine learning model can receive the amount of charge [SoC(t)], temperature [T(t)], the first-order change in temperature [T(t)-T(t-1)], and the second-order change in temperature [T(t)+T(t-2)-2T(t-1)] as input profiles. For example, the two-dimensional change in temperature can have physical information that can detect a sudden change in current that has a large impact on the temperature. In addition, the amount of charge [SoC(t)] can be obtained as the accumulated value of the current over time.

[0097] A machine learning model (S125) according to one embodiment of the present invention is a current that can satisfy the input charge amount and temperature [ ] profile can be predicted as an output value. For example, the predicted current profile can function as a charge / discharge strategy that can satisfy a given charge amount and temperature profile.

[0098] For example, the processor (150) can obtain a current profile for the temperature profile while knowing the temperature profile [T(t)] and the initial state of charge [SoC(0)] among the variables used as input values.

[0099] In addition, the processor (150) can utilize the previously secured temperature profile to secure the first and second temperature variations of Equation 4. For example, the state of charge at the next point in time can be updated through the predicted current as in Equation 6.

[0100] <Mathematical Formula 6>

[0101]

[0102] Here, the reason for multiplying by 50 / 3600 is because the unit of current is C-rate, so it is charged / discharged by the amount for 50 seconds in 1 hour.

[0103] The processor (150) determines the current at each point in time according to the time index (t). ] can be predicted. For example, the processor (150) can predict the final current profile [ for all time indices (t=1,2,..,72). ] can be secured.

[0104] According to one embodiment, the step of verifying and retraining the machine learning model (S130, S140) may include a step of predicting a predicted value for the current of the battery based on the temperature of the battery through the machine learning model using the verification data that is not included in the training data, a step of comparing the predicted value for the current with the current data of the verification data, and a step of comparing the temperature data of the battery obtained from the predicted value for the current with the temperature data of the battery of the verification data.

[0105] For example, the processor (150) uses the verification data (S112) to predict the charge / discharge profile [ ] can verify the authenticity. For example, the verification method is 1) comparing the current profile [I(t)] that causes the input profile used for prediction and the predicted current profile [ ] or 1) comparing the difference between the predicted current profiles, or 2) re-acquiring the temperature profile [T'(t)] using the method of step S110 with the predicted current profile and comparing it with the original temperature profile.

[0106] For example, the processor (150) can obtain verification data, which is a true value, through the method of step S110. For example, the processor (150) can obtain verification data including a current profile and a temperature profile.

[0107] Next, the processor (150) can predict a predicted value for the battery current based on the battery temperature using a machine learning model using verification data not included in the training data. Subsequently, the processor (150) can compare the predicted value for the current with the current data of the verification data.

[0108] Additionally, the processor (150) can compare the temperature data of the battery obtained from the predicted value for current with the temperature data of the battery in the verification data.

[0109] According to one embodiment of the present invention, the processor (150) can derive a charge / discharge strategy using the machine learning model if the convergence of the machine learning model is secured using the method 1) or 2) above. Alternatively, according to one embodiment of the present invention, the processor (150) can retrain the machine learning model if the convergence of the machine learning model is insufficient. For example, the convergence of the machine learning model can be determined to be achieved if the error between the comparison data falls within a preset range.

[0110] For example, if the convergence of the machine learning model is not secured, the processor (150) can additionally acquire training data and verification data. For example, the processor (150) can additionally acquire new training data and, if necessary, verification data using the same method as in step S110. For example, in the case of verification data, if the number of training data samples is confirmed to be lower than the previously maintained ratio (1 to 5%), additional data can be acquired. Subsequently, the processor (150) can re-perform training and verification of the machine learning model using the enhanced training data and verification data.

[0111] According to one embodiment, the step of predicting a current profile according to a temperature profile of the battery may include a step of predicting a change in charge / discharge current of the battery according to time that satisfies a change in temperature of the battery according to time.

[0112] For example, if the processor (150) secures convergence of the machine learning model through verification, the learned machine learning model can be ultimately stored and utilized to derive a battery charge / discharge strategy in the future. For example, the learned machine learning model can derive a current profile (i.e., a charge / discharge strategy) that satisfies a given temperature profile by reversely predicting it.

[0113] FIG. 4 and FIG. 5 are drawings for explaining a battery charge / discharge current prediction method according to one embodiment of the present invention.

[0114] Referring to FIGS. 4 and 5 , a battery current profile, a battery temperature profile, a current profile predicted from the battery temperature profile, and a temperature profile recalculated from the predicted current profile are illustrated according to an embodiment of the present invention. Here, the battery temperature profile can be derived from the current profile through the method of step S110. For example, in the embodiment of the present invention, a total of 4,397 training data and 77 verification data were used.

[0115] For example, referring to FIGS. 4(a) to 4(d), examples of verification data are illustrated. A machine learning model can be verified by comparing a current profile predicted from a battery current profile and a battery temperature profile. Alternatively, a machine learning model can be verified by comparing a temperature profile recalculated from a battery temperature profile and a predicted current profile. Here, the horizontal axis represents time (unit: seconds), and the vertical axis can represent current intensity (unit: C-rate) for current and internal battery temperature (unit: K) for temperature.

[0116] For example, referring to FIG. 5(a) and FIG. 5(b), examples of verification data are shown in which a charge / discharge strategy different from the true value was obtained by predicting it through a machine learning model, but the recalculated temperature is similar to the true value.

[0117] Current profile predicted by the technique proposed in the present invention [ ] may be different from the true value [I(t)]. However, referring to FIG. 5, the temperature profile [T'(t)] re-acquired through the method of step S110 may show the same result as the true value [T(t)]. Based on this, the technique proposed in the present invention can suggest an efficient battery usage method to the user by suggesting not only a simple reverse prediction of the charge / discharge strategy, but also a multi-faceted charge / discharge strategy.

[0118] The present invention is a method for predicting charge and discharge information that enables a battery to produce an arbitrary temperature output by learning temperature data that can be obtained during a battery operation process.

[0119] Unlike forward prediction that predicts the current state (charge amount, etc.) from battery output data, the present invention can derive a battery charge / discharge strategy that satisfies arbitrary battery usage restrictions (temperature, etc.) through reverse prediction.

[0120] Unlike predicting future usage patterns based on users' actual usage patterns, the present invention can provide users with a real-time battery usage strategy based on current usage by establishing a strategy that can operate the battery within a normal range.

[0121] Unlike conventional techniques that required additional optimization procedures for establishing battery usage strategies, the present invention can immediately predict battery usage strategies through reverse prediction of a learning model.

[0122] By utilizing the present invention, a machine learning-based reverse prediction model can be obtained that predicts a charge / discharge profile that satisfies a given battery temperature profile, and in the process, the amount of learning data is sequentially increased, thereby improving learning efficiency.

[0123] According to an embodiment of the present invention, a charge / discharge profile can be directly predicted from a given battery temperature profile, thereby quickly and accurately proposing a charge / discharge strategy required for real-time dynamic battery thermal management.

[0124] According to an embodiment of the present invention, the obtained prediction model can contribute to establishing various charge / discharge strategies by suggesting not only the original charge / discharge profile that caused the battery temperature profile, but also an alternative charge / discharge profile.

[0125] Although the present invention has been described above with reference to limited embodiments and drawings, the present invention is not limited thereto, and it is obvious that various modifications and variations are possible within the scope of the technical idea of ​​the present invention and the equivalent scope of the patent claims to be described below by a person having ordinary skill in the art to which the present invention pertains.

[0126] The various embodiments described above may be implemented in the form of a computer program that can be executed through various components on a computer, and such a computer program may be recorded on a computer-readable medium. In this case, the medium may be one that continuously stores the computer-executable program, or one that temporarily stores it for execution or download. In addition, the medium may be various recording or storage means in the form of a single or multiple hardware combinations, and is not limited to a medium directly connected to a computer system, but may also be distributed over a network. Examples of the medium may include magnetic media such as hard disks, floppy disks, and magnetic tapes, optical recording media such as CD-ROMs and DVDs, magneto-optical media such as floptical disks, and those configured to store program instructions, including ROM, RAM, and flash memory. In addition, examples of other media may include recording or storage media managed by app stores that distribute applications, sites that supply or distribute various software, servers, etc.

[0127] In this specification, "part", "module", etc. may be a hardware component such as a processor or a circuit, and / or a software component executed by a hardware component such as a processor. For example, "part", "module", etc. may be implemented by components such as software components, object-oriented software components, class components, and task components, and processes, functions, properties, procedures, subroutines, segments of program code, drivers, firmware, microcode, circuits, data, databases, data structures, tables, arrays, and variables.

Claims

1. A method for predicting battery charging and discharging current performed by a computing device, A step of obtaining learning data and verification data including a temperature profile according to a charge / discharge current profile of a battery; A step of training a machine learning model using the above training data; A step of verifying and re-learning the machine learning model learned using the above verification data; and A step of predicting a current profile according to a temperature profile of a battery using the above machine learning model; A method for predicting battery charge / discharge current, comprising:

2. In paragraph 1, A method for predicting a battery charge / discharge current, wherein the step of predicting a current profile according to a temperature profile of the battery includes a step of predicting a change in the charge / discharge current of the battery according to time that satisfies a change in the temperature of the battery according to time.

3. In paragraph 1, A method for predicting battery charge / discharge current, wherein the step of acquiring the above learning data and verification data includes the step of acquiring a plurality of data on changes in the charge / discharge current of the battery over time and changes in the temperature of the battery according to changes in the charge / discharge current of the battery.

4. In paragraph 3, A method for predicting battery charge / discharge current, wherein the step of obtaining the above learning data and verification data includes the step of obtaining a plurality of data by random sampling that extracts a variable combination having a homogeneous distribution within a given variable combination and range.

5. In paragraph 4, A method for predicting battery charge / discharge current, wherein the step of acquiring the learning data and the verification data includes the step of acquiring the learning data and the verification data, each of which has the same distribution but includes different current and temperature data.

6. In paragraph 1, The step of learning the above machine learning model is to learn the temperature of the battery over time [T(t)] and the first order change in the temperature of the battery [ T(t)], the second order change in the temperature of the battery[ A method for predicting battery charge and discharge current, comprising the step of learning the machine learning model that receives as input values ​​the state of charge of the battery [SoC(t)] over time and the current of the battery over time as output values.

7. In paragraph 6, The first order change in temperature of the above battery[ T(t)] and the second order change in temperature of the battery [ T(t)] has the following relationship with the temperature of the battery over time [T(t)]: 1st order change: 2nd order change: , a method for predicting battery charge and discharge current.

8. In paragraph 6, The state of charge of the battery over time [SoC(t)] is the predicted current of the battery over time [ ] is expressed as the following formula based on , a method for predicting battery charge and discharge current.

9. In paragraph 1, The steps of verifying and retraining the above machine learning model are: A step of predicting a prediction value for the current of a battery based on the temperature of the battery through the machine learning model using the verification data that is not included in the training data; A step of comparing the predicted value for the above current with the current data of the above verification data; and A method for predicting battery charge / discharge current, comprising the step of comparing battery temperature data obtained from a predicted value for the current with battery temperature data of the verification data.

10. A computer program stored in a recording medium for executing the method of any one of claims 1 to 9 using a computing device.

11. Memory for storing learning data and verification data including a temperature profile according to the charge / discharge current profile of the battery; and A battery charge / discharge current prediction device comprising at least one processor that obtains the learning data and the verification data, learns a machine learning model using the learning data, verifies and re-learns the learned machine learning model using the verification data, and predicts a current profile according to a temperature profile of the battery using the machine learning model.

12. In paragraph 11, The above processor is a battery charge / discharge current prediction device that predicts the change in charge / discharge current of a battery over time that satisfies the change in temperature of the battery over time.

13. In paragraph 11, The above processor is a battery charge / discharge current prediction device that obtains a plurality of data on changes in the charge / discharge current of the battery over time and changes in the temperature of the battery according to changes in the charge / discharge current of the battery.

14. In paragraph 11, The above processor calculates the temperature of the battery over time [T(t)], the first order change in the temperature of the battery [ T(t)], the second order change in the temperature of the battery[ A battery charge / discharge current prediction device that learns the machine learning model that receives the state of charge of the battery [SoC(t)] over time and the state of charge of the battery over time as input values ​​and outputs the current of the battery over time as an output value.

15. In paragraph 11, The above processor predicts a predicted value for the current of the battery based on the temperature of the battery through the machine learning model using the verification data not included in the learning data, compares the predicted value for the current with the current data of the verification data, and compares the temperature data of the battery obtained from the predicted value for the current with the temperature data of the battery of the verification data. A battery charge / discharge current prediction device.

Images