Device and method for estimating battery aging state

Abstract

One embodiment provides a method for estimating the aging state of a battery, comprising: a step of storing a charging time measured at each of a plurality of terminal voltage intervals as a reference charging time for a battery in a reference aging state that is charged with constant current; a step of storing a charging time for each of N (N is a natural number greater than or equal to 2) terminal voltage intervals as a comparison charging time in a constant current charging mode of the battery, and calculating a ratio value between the reference charging time and the comparison charging time for each of the N terminal voltage intervals; and a step of estimating the aging state of the battery by inputting the N ratio values ​​into a pre-trained machine learning model.

Description

Technology Field

[0001] This embodiment relates to a technology for estimating the aging state of a battery. Background Technology

[0002] Known performance degradation mechanisms for lithium-based batteries include LLI (Loss of Lithium Inventory), LAM (Loss of Active Material), ORI (Ohmic Resistance Increase), and lithium plating.

[0003] LLI is a major cause of battery capacity reduction and refers to the loss of available lithium ions, and is known to be caused mainly by the continuous growth of the SEI (Solid Electrolyte Interface) layer.

[0004] LAM refers to the structural and mechanical degradation of electrodes, and it can reduce both battery capacity and output. While LLI can occur independently, LAM is known to occur simultaneously with LLI.

[0005] ORI refers to a phenomenon where the performance degradation of electrode and electrolyte materials manifests as an increase in the cell's electron and ion resistance, and it can be caused by various factors such as LLI and LAM.

[0006] Lithium plating refers to a mechanism that causes LAM by blocking the pores of porous materials or splitting the collector into thin layers. Lithium plating not only increases the rate of cell capacity degradation but can also cause dendrite growth, leading to internal short circuits in the cell.

[0007] A state in which performance is degraded by such a mechanism can be called a state of aging, and a state of aging is also denoted as SOH (state-of-health).

[0008] Known methods for estimating SOH include incremental capacity analysis and methods utilizing constant voltage mode charging current characteristics.

[0009] However, the incremental capacity analysis method has the disadvantage that it is time-consuming because it requires obtaining data through many cycles of full buffering or full discharge with a low current corresponding to 1 / 25C, and it cannot avoid the problem of data measurement noise that is inevitably involved.

[0010] Furthermore, the method utilizing constant voltage mode charging current characteristics has the disadvantage of being difficult to apply to actual products because it requires a significant amount of time to collect data—as it necessitates completing constant current mode charging and measuring the charging current in constant voltage mode—and because in actual field applications, only partial charging is often performed in constant current mode. The problem to be solved

[0011] Against this backdrop, the objective of the present embodiment is, in one aspect, to provide a technology for estimating an aging state that is easy to apply to actual products. In another aspect, the objective of the present embodiment is to provide a technology that can reduce the time required to estimate the aging state. means of solving the problem

[0012] To achieve the aforementioned objective, one embodiment provides a method for estimating the aging state of a battery, comprising: a step of storing a charging time measured at each of a plurality of terminal voltage intervals as a reference charging time for a battery in a reference aging state that is charged with constant current; a step of storing a charging time for each of N (N is a natural number greater than or equal to 2) terminal voltage intervals as a comparison charging time in a constant current charging mode of the battery, and calculating a ratio value between the reference charging time and the comparison charging time for each of the N terminal voltage intervals; and a step of estimating the aging state of the battery by inputting the N ratio values ​​into a pre-trained machine learning model.

[0013] In the step of estimating the aging state, the above method can estimate the aging state of the battery by further inputting the difference between the operating temperature and the reference temperature of the battery into the machine learning model.

[0014] The size of each of the above N terminal voltage intervals may be the same.

[0015] The above method can estimate the aging state of the battery by additionally inputting at least one terminal voltage value corresponding to the N terminal voltage intervals into the machine learning model in the step of estimating the aging state.

[0016] In the step of calculating the ratio value, the above method may set N intervals from the terminal voltage of the battery confirmed at one point in time as the N terminal voltage intervals.

[0017] In the above method, when the comparison charging time is measured for M (where M is a natural number greater than N) terminal voltage intervals, in the step of calculating the ratio value, the N terminal voltage intervals can be set such that predetermined terminal voltage intervals are included.

[0018] Among the M terminal voltage intervals, the charging time in one terminal voltage interval belonging to the N terminal voltage intervals may be longer than the charging time in another terminal voltage interval not belonging to the N terminal voltage intervals.

[0019] The above machine learning model may be in the form of an ensemble of N sub-machine learning models that take the ratio value of each terminal voltage range as input and the aging state value as output.

[0020] The above machine learning model is composed of L (where L is a natural number greater than N) sub-machine learning models, each having a ratio value of each terminal voltage range as input and an aging state value as output, and each sub-machine learning model has a set priority, and the above method can estimate the aging state of the battery according to the output value of the sub-machine learning model with a higher priority among the output values ​​of the N sub-machine learning models corresponding to the N terminal voltage ranges among the L sub-machine learning models in the step of estimating the aging state.

[0021] Among the above N terminal voltage intervals, the sizes of at least two terminal voltage intervals are different from each other, and in the above at least two terminal voltage intervals, the reference charging time or the comparison charging time may have similar sizes within a certain error range.

[0022] Another embodiment provides a battery aging state estimation device comprising: a storage circuit that stores a charging time measured at each of a plurality of terminal voltage intervals as a reference charging time for a reference aging state battery that is charged with constant current; a calculation circuit that stores a charging time for each of N (N is a natural number greater than or equal to 2) terminal voltage intervals as a comparison charging time in a constant current charging mode of a battery, and calculates a ratio value between the reference charging time and the comparison charging time for each of the N terminal voltage intervals; and a state estimation circuit that estimates the aging state of the battery by inputting the N ratio values ​​into a pre-trained machine learning model.

[0023] The above state estimation circuit can estimate the aging state of the battery by further inputting the difference between the operating temperature and the reference temperature of the battery into the machine learning model.

[0024] The above state estimation circuit can estimate the aging state of the battery by further substituting at least one terminal voltage value corresponding to the N terminal voltage intervals into the machine learning model.

[0025] The above calculation circuit can set N intervals from the terminal voltage of the above battery confirmed at one point in time as the N terminal voltage intervals.

[0026] When the comparison charging time is measured for M (where M is a natural number greater than N) terminal voltage intervals, the calculation circuit can set the N terminal voltage intervals such that predetermined terminal voltage intervals are included.

[0027] Another embodiment provides a method for estimating the aging state of a battery, comprising: a step of calculating the time rate of change of terminal voltage in each of N (N is a natural number greater than or equal to 2) terminal voltage intervals in a constant current charging mode of a battery; and a step of estimating the aging state of the battery by comparing the calculated N terminal voltage time rates of change with terminal voltage time rate of change data stored according to the aging state.

[0028] The above terminal voltage time change rate data may include a lookup table that stores the terminal voltage time change rates in multiple terminal voltage intervals according to the aging state.

[0029] In the step of estimating the aging state, the above method can find data storing terminal voltage time change rates that have a high similarity to the calculated N terminal voltage time change rates in the lookup table, and determine the aging state value corresponding to the data as the aging state value of the battery.

[0030] In the step of estimating the aging state, the above method calculates a Euclidean distance value for the N calculated terminal voltage time change rates and the terminal voltage time change rates stored in a lookup table, finds the data with the smallest Euclidean distance value, and determines the aging state value corresponding to the data as the aging state value of the battery.

[0031] The range of the plurality of terminal voltage ranges corresponding to the above lookup table may be wider than the range of the N terminal voltage ranges calculated for the above battery.

[0032] The above lookup table is stored by temperature, and in the step of estimating the aging state, the method selects a lookup table corresponding to the temperature value of the one battery among the lookup tables included in the terminal voltage time change rate data, and the aging state value of the one battery can be determined according to the selected lookup table.

[0033] At least two of the above N terminal voltage intervals may have different sizes.

[0034] Among the above at least two terminal voltage intervals, the size of the terminal voltage interval with a larger terminal voltage time change rate may be smaller than the size of the terminal voltage interval with a smaller terminal voltage time change rate.

[0035] The above terminal voltage time change rate data may be data measured for each charge / discharge cycle for the above-mentioned battery and batteries of the same type.

[0036] Another embodiment includes a measurement circuit that records terminal voltage values ​​of a battery while counting the charging time in a constant current charging mode of a battery;

[0037] A battery aging state estimation device is provided, comprising: a storage circuit that stores terminal voltage time change rate data recording terminal voltage time change rates in multiple terminal voltage intervals according to aging state; and a state estimation circuit that calculates the terminal voltage time change rate in each of N (N is a natural number greater than or equal to 2) terminal voltage intervals using the charging time and the terminal voltage values, and estimates the aging state of the battery by comparing the calculated N terminal voltage time change rates with the terminal voltage time change rate data.

[0038] The above terminal voltage time change rate data may include a lookup table that stores the terminal voltage time change rates in multiple terminal voltage intervals according to the aging state.

[0039] The above state estimation circuit can find data storing terminal voltage time change rates that have a high similarity to the calculated N terminal voltage time change rates in the lookup table, and determine the aging state value corresponding to the data as the aging state value of the battery.

[0040] The above state estimation circuit can calculate a Euclidean distance value for the N calculated terminal voltage time change rates and the terminal voltage time change rates stored in the lookup table, and determine the aging state value corresponding to the data with the smallest Euclidean distance value as the aging state value of the battery.

[0041] The above lookup table is stored by temperature, and the state estimation circuit selects a lookup table corresponding to the temperature value of the battery among the lookup tables included in the terminal voltage time change rate data, and can determine the aging state value of the battery according to the selected lookup table.

[0042] At least two of the above N terminal voltage intervals may have different sizes. Effects of the invention

[0043] As described above, according to the present embodiment, a highly accurate aging state estimation technology can be easily applied to actual products. Furthermore, according to the present embodiment, the time required to estimate the aging state can be reduced. Brief explanation of the drawing

[0044] FIG. 1 is a configuration diagram of a battery charging system according to an embodiment of the present specification. Figure 2 is a graph showing that the constant current charging time in a specific terminal voltage range varies according to SOH. Figure 3 is a diagram showing a method for estimating SOH based on a specific charging time. FIG. 4 is a configuration diagram of a battery aging state estimation device according to the first embodiment. FIG. 5 is a first example configuration diagram of a machine learning model according to the first embodiment. Figure 6 is a diagram showing the terminal voltage range of reference data and the terminal voltage range of the vector input to the machine learning model. Figure 7 is a diagram showing the change in charging time according to the terminal voltage range. Figure 8 is a diagram illustrating an example of selecting a terminal voltage range to be input into a machine learning model from a terminal voltage range where the comparison charging time is measured. FIG. 9 is a second example configuration diagram of a machine learning model according to the first embodiment. FIG. 10 is a third example configuration diagram of a machine learning model according to the first embodiment. FIG. 11 is a diagram of the fourth example configuration of a machine learning model according to the first embodiment. FIG. 12 is a diagram of the fifth example configuration of a machine learning model according to the first embodiment. FIG. 13 is a flowchart of a method for estimating battery aging state according to the first embodiment. FIG. 14 is a flowchart of a method for estimating battery aging state according to a second embodiment. FIG. 15 is an example drawing of a lookup table that stores the terminal voltage time change rate for each charge / discharge cycle in the second embodiment. FIG. 16 is an example drawing of lookup tables stored by temperature in the second embodiment. FIG. 17 is a diagram illustrating the process of finding the data most similar to the terminal voltage time change rate of the target battery in the second embodiment. FIG. 18 is a flowchart of a method for estimating battery aging state according to a second embodiment. Specific details for implementing the invention

[0045] Hereinafter, some embodiments of the present invention will be described in detail with reference to the exemplary drawings. It should be noted that in assigning reference numerals to the components of each drawing, the same components are given the same reference numeral whenever possible, even if they are shown in different drawings. Furthermore, in describing the present invention, if it is determined that a detailed description of related known components or functions could obscure the essence of the invention, such detailed description is omitted.

[0046] In addition, terms such as first, second, A, B, (a), (b), etc., may be used when describing the components of the present invention. These terms are intended only to distinguish the components from other components, and the essence, order, or sequence of the components is not limited by the terms. Where it is stated that a component is "connected," "combined," or "connected" to another component, it should be understood that the component may be directly connected or connected to the other component, but that another component may also be "connected," "combined," or "connected" between each component.

[0047] FIG. 1 is a configuration diagram of a battery charging system according to an embodiment of the present specification.

[0048] Referring to FIG. 1, the battery charging system (100) may include a battery (120), a battery management device (110), and a charger (130), etc.

[0049] The battery (120) may be a lithium-based battery. The battery (120) may be a lithium-ion battery and may be a lithium-polymer battery.

[0050] The battery (120) may be composed of a single cell or a plurality of cells. When the battery (120) is composed of a plurality of cells, each cell may be connected in series or in parallel.

[0051] The battery (120) may be configured in the form of a battery pack containing cells. In addition to the cells, the battery pack may further include a protection circuit and sensors such as a temperature sensor (TS).

[0052] The battery (120) may further include a signal transmission and reception circuit. The battery (120) may transmit status information of the battery (120) to an external device through the signal transmission and reception circuit. For example, the battery (120) may transmit information (TV) regarding the operating temperature of the battery (120) to the battery management device (110) through the signal transmission and reception circuit.

[0053] The battery (120) may have terminals formed therein for input and output of current. For example, the battery (120) may have a positive terminal representing positive polarity and a negative terminal representing negative polarity. The voltage between the two terminals formed in the battery (120) may be referred to as terminal voltage (VT).

[0054] The current flowing into or out of the battery (120) can be referred to as battery current (IB). The current flowing into the positive terminal of the battery (120) can be referred to as charging current, and the current flowing out of the positive terminal of the battery (120) can be referred to as discharging current.

[0055] The battery (120) and the charger (130) may be connected by a cable through which battery current (IB) can flow. The charger (130) can supply charging current to the battery (120) through this cable.

[0056] A current sensor (IS) may be placed on the cable. The battery management device (110) can measure the magnitude of the battery current (IB) through the current sensor (IS). Additionally, the battery management device (110) can measure the terminal voltage (VT) of the battery (120) using the voltage formed on the cable.

[0057] The charger (130) can operate in a constant current charging mode, a constant voltage charging mode, and a constant power charging mode. In the constant current charging mode, the charger (130) can make the magnitude of the charging current supplied to the battery (120) constant. For example, when the magnitude of the charging current that can fully charge the battery (120) in one hour is 1C, the charger (130) can maintain the magnitude of the charging current at 0.1C, at 0.2C, or at other constant magnitudes in the constant current charging mode.

[0058] In a constant voltage charging mode, the charger (130) can supply a charging current so that the terminal voltage (VT) of the battery (120) is maintained constant, and in a constant power charging mode, the charger (130) can supply a charging current so that the magnitude of the charging power supplied to the battery (120) is maintained constant.

[0059] The battery management device (110) can estimate the state of the battery (120) and control a peripheral device—e.g., a charger (130)—for the management of the battery (120).

[0060] The battery management device (110) can estimate the charge state of the battery (120). The charge state can be expressed as SOC (state-of-charge). The battery management device (110) can estimate the SOC of the battery (120) using the charge integration method, the OCV (open circuit voltage) method, etc.

[0061] The battery management device (110) can estimate the remaining capacity of the battery (120). The battery management device (110) can estimate the remaining capacity of the battery (120) by summing the amount of charge discharged until the fully charged battery (120) reaches a fully discharged state.

[0062] The battery management device (110) can estimate the aging state of the battery (120). The aging state can be expressed as SOH (state-of-health). In terms of estimating the aging state of the battery (120), the battery management device (110) can be referred to as a battery aging state estimation device.

[0063] The following description focuses on an embodiment in which the battery management device (110) functions as a battery aging state estimation device, but as previously mentioned, the battery management device (110) can estimate the charge state of the battery, estimate the remaining capacity of the battery, and perform other functions related to battery management.

[0064] SOH can be calculated as the ratio of the battery capacity in the reference state to the battery capacity in the current state.

[0065] SOH = Ck / Ci

[0066] Ck: Battery capacity in current state, Ci: Battery capacity in reference state

[0067] Battery capacity can be defined as the total amount of charge being discharged or charged.

[0068]

[0069] I : Battery current

[0070] t k : Time point 1, t k+L : A second point in time after a certain amount of time has elapsed since the first point in time

[0071] SOC k : Battery SOC at the first point in time, SOC k+L : Battery SOC at point 2

[0072] Voc k : Battery open-circuit voltage at the first point in time, Voc k+L : Battery open-circuit voltage at the second point in time

[0073] Generally, battery SOC maps 1:1 to the open-circuit voltage Voc, and since the charging current of a battery charged in constant current mode is constant, the battery capacity Ck of a given battery terminal voltage range is given by the constant current charging time t k+L - t k It can be proportional to. Therefore, the SOH of the battery can be calculated using the following modified formula.

[0074] SOH = Ck / Ci ≈ Δtk / Δti

[0075] Δtk: Constant current charging time for a specific terminal voltage range in the current state

[0076] Δti: Constant current charging time for a specific terminal voltage range in the reference state

[0077] Figure 2 is a graph showing that the constant current charging time in a specific terminal voltage range varies according to SOH.

[0078] Referring to Figure 2, it can be seen that when the initial cell (Cycle #1) is charged in constant current mode, it takes a first charging time (Δt1) in a specific terminal voltage range (ΔV), the cell that has been charged and discharged 100 times (Cycle #100) takes a second charging time (Δt2), the cell that has been charged and discharged 200 times (Cycle #200) takes a third charging time (Δt3), the cell that has been charged and discharged 300 times (Cycle #300) takes a fourth charging time (Δt4), and the cell corresponding to SOH 70% (Cycle #364) takes a fifth charging time (Δt5).

[0079] Generally, considering that SOH decreases as the number of charge / discharge cycles increases, it can be seen through Figure 2 that the charging time for a specific terminal voltage range (ΔV) becomes shorter as SOH decreases.

[0080] As shown in FIG. 2, the battery aging state estimation device can determine a specific terminal voltage range (ΔV), check the battery charging time of a reference state—e.g., initial state—corresponding to the terminal voltage range (ΔV) and the battery charging time of a current state, and estimate the SOH by calculating the ratio value of the battery charging time of the reference state and the battery charging time of the current state.

[0081] As another method, the battery aging state estimation device can determine a specific charging time in constant current charging mode and then determine the terminal voltage range for that charging time.

[0082] Figure 3 is a diagram showing a method for estimating SOH based on a specific charging time.

[0083] Referring to FIG. 3, the battery aging state estimation device can first determine a specific charging time (Δtk) when the battery in the current state is charging at a constant current. Then, the battery aging state estimation device can check the terminal voltages (V1, V2) at the start time (t1) and end time (t2) of the corresponding charging time (Δtk) and search for the corresponding terminal voltage range (ΔV) in the previously stored reference state battery data. Then, the battery aging state estimation device can determine the charging time (Δti) of the reference state battery according to the charging start time (t1') and charging end time (t2') of the reference state battery corresponding to the corresponding terminal voltage range (ΔV). Then, the battery aging state estimation device can estimate the SOH by calculating the ratio value between the reference state battery charging time and the current state battery charging time.

[0084] When the terminal voltage range is determined first, the magnitude of the charging time may be small in a specific terminal voltage range, and when the magnitude of the charging time is small, the error in SOH estimation may be large. However, when the charging time is determined first as in the method shown in Fig. 3, there is an advantage in that the possibility of such error occurring can be reduced.

[0085] FIG. 4 is a configuration diagram of a battery aging state estimation device according to the first embodiment.

[0086] Referring to FIG. 4, a battery aging state estimation device (hereinafter referred to as the 'estimation device', 400) may include a storage circuit (410), a calculation circuit (420), and a state estimation circuit (430), etc.

[0087] The storage circuit (410) can store the charging time measured at each of the multiple terminal voltage intervals as the reference charging time for a battery in a reference aging state that is charged with constant current.

[0088] The reference aging state may be a battery in an initial state. A device for generating reference data may prepare multiple batteries in an initial state, measure the reference charging time for each of the multiple batteries, and then perform statistical processing—e.g., calculating the average—to determine the final reference charging time. A storage circuit (410) may store the reference charging time determined in this way in memory.

[0089] The reference charging time can be measured at a reference temperature. For example, the reference data generation device can set the reference temperature to 20 degrees Celsius and measure and record the reference charging time at each of the multiple terminal voltage intervals for a battery that is charged with constant current at 20 degrees Celsius. Then, the storage circuit (410) can store the reference charging time measured at the reference temperature in memory.

[0090] And, the storage circuit (410) can store data regarding the reference charging time recorded by the reference data generation device in memory.

[0091] The calculation circuit (420) can store the charging time for each of N (N is a natural number greater than or equal to 2) terminal voltage intervals as a comparison charging time in the constant current charging mode of the battery (hereinafter referred to as the 'target battery') that is the subject of state estimation. Additionally, the calculation circuit (420) can calculate the ratio value of the reference charging time and the comparison charging time for each of the N terminal voltage intervals.

[0092] The state estimation circuit (430) can estimate the aging state of the target battery by inputting N ratio values ​​into a pre-trained machine learning model.

[0093] FIG. 5 is a first example configuration diagram of a machine learning model according to the first embodiment.

[0094] Referring to FIG. 5, N ratio values ​​(R1, R2, ..., Rn) can be input to the machine learning model (510).

[0095] The estimation device can calculate a first ratio value (R1) of a reference charging time (Δti) and a comparison charging time (Δtk) for a first terminal voltage section, calculate a second ratio value (R2) for a second terminal voltage section that is continuous with the first terminal voltage section, and sequentially calculate an Nth ratio value (Rn) for the Nth terminal voltage section.

[0096] And, the estimation device can form a vector (VEC) with the first ratio value (R1) to the Nth ratio value (Rn) and input the vector (VEC) into a machine learning model (510).

[0097] And, the estimation device can estimate the SOH output from the machine learning model (510) as the aging state of the target battery.

[0098] The estimation device can generate a vector (VEC) value for N consecutive terminal voltage intervals in the target battery. The terminal voltage range for which this vector (VEC) is generated may be narrower than the terminal voltage range stored in the reference data.

[0099] Figure 6 is a diagram showing the terminal voltage range of reference data and the terminal voltage range of the vector input to the machine learning model.

[0100] Referring to Fig. 6, the terminal voltage range of the reference charging time stored as reference data may be wider than the terminal voltage range of the vector input to the machine learning model.

[0101] For example, the terminal voltage range of the reference data may be in the range of 3.0V to 4.2V, and the terminal voltage range of the vector input to the machine learning model may correspond to a part of the range of 3.0V to 4.2V.

[0102] The terminal voltage range of a vector input to a machine learning model can vary within the terminal voltage range of the reference data. For example, the terminal voltage range of a vector may correspond to a relatively high voltage range, such as the first vector (VEC1), a relatively low voltage range, such as the third vector (VEC3), or an intermediate voltage range, such as the second vector (VEC2).

[0103] The estimation device can estimate the aging state of the target battery with relatively high accuracy even if the terminal voltage range of the vector changes. For example, even if the terminal voltage range of the vector input to the machine learning model corresponds to the terminal voltage range of the first vector (VEC1) or the terminal voltage range of the third vector (VEC3), the result may be similar. This is because the input to the machine learning model is not the charging time of the terminal voltage range, but the ratio of the reference charging time to the comparison charging time within the terminal voltage range.

[0104] The target battery can be charged with constant current at different voltage ranges depending on the situation. Since the estimation device according to the first embodiment can estimate the aging state of the target battery using a vector (VEC) obtained at different voltage ranges, it has the advantage of being applicable to various situations and various applications.

[0105] Figure 7 is a diagram showing the change in charging time according to the terminal voltage range.

[0106] Referring to Fig. 7, it can be seen that as the number of charge and discharge cycles increases, the charging time for the same terminal voltage range becomes shorter.

[0107] Meanwhile, when examining the entire terminal voltage range where the reference charging time is recorded by dividing it into three ranges (710, 720, 730), it can be seen that the second range (720) has a relatively longer charging time per unit voltage compared to the other ranges (710, 730).

[0108] In sections where the length of the charging time per unit voltage is relatively long, such as in the second range (720), even if an error occurs in the charging time due to noise or environmental influence, the variation in the value of the estimated aging state may be small.

[0109] Therefore, when the estimation device can select multiple ranges to estimate a more stable and accurate aging state, it can estimate the aging state by selecting a range that has been previously confirmed to have high stability and accuracy, such as the second range (720).

[0110] Figure 8 is a diagram illustrating an example of selecting a terminal voltage range to be input into a machine learning model from a terminal voltage range where the comparison charging time is measured.

[0111] Referring to FIG. 8, the estimation device can measure the comparative charging time for a target battery for a relatively large terminal voltage range (810). At this time, the measured terminal voltage range (810) may be a voltage range larger than the terminal voltage range to be input into a machine learning model.

[0112] When inputting a voltage range corresponding to N terminal voltage intervals into a machine learning model, the measured terminal voltage range (810) may include M terminal voltage intervals greater than N.

[0113] The estimation device can select vectors (VEC1, VEC2, VEC3, VEC4) with various voltage ranges from M terminal voltage intervals, and the estimation device can set N terminal voltage intervals such that they include predetermined terminal voltage intervals—e.g., terminal voltage intervals corresponding to VEC3—and generate ratio values ​​and / or vectors (VEC3) in the corresponding terminal voltage intervals.

[0114] As explained in Fig. 7, among the measured M terminal voltage intervals, the charging time in one terminal voltage interval belonging to the predetermined N terminal voltage intervals may be longer than the charging time in another terminal voltage interval not belonging to the N terminal voltage intervals.

[0115] The size of each terminal voltage interval may be the same. Alternatively, the sizes of at least two of the N terminal voltage intervals may differ from each other, in which case the reference charging time or comparison charging time in at least two terminal voltage intervals may be the same or have similar sizes within a certain error range. This embodiment can be understood as applying the concept described with reference to FIG. 3.

[0116] Conceptually, the method of estimating the aging state by calculating ratio values ​​has the advantage of allowing the range of terminal voltage to be freely selected compared to methods that do not. However, since there may be slight variations in the estimated value depending on the range of terminal voltage, additional terminal voltage values ​​can be input into the machine learning model to compensate for this.

[0117] FIG. 9 is a second example configuration diagram of a machine learning model according to the first embodiment.

[0118] Referring to FIG. 9, a vector (VEC) having N ratio values ​​(R1, R2, ..., Rn) can be input to the machine learning model (910), and at least one terminal voltage value (V1, V2) corresponding to N terminal voltage intervals can be further input.

[0119] The machine learning model (910) can receive the terminal voltages (V1, V2) of N terminal voltage intervals, and the lowest voltage (V1) or the highest voltage (V2) can be received.

[0120] This method can further improve the accuracy of the aging state estimation by inputting more information about the voltage ranges of the terminal voltage intervals input to the machine learning model (910).

[0121] FIG. 10 is a third example configuration diagram of a machine learning model according to the first embodiment.

[0122] Referring to FIG. 10, a vector (VEC) having N ratio values ​​(R1, R2, ..., Rn) can be input to the machine learning model (1010), at least one terminal voltage value (V1, V2) corresponding to N terminal voltage intervals can be input, and the difference (ΔT) between the operating temperature (Tr) of the target battery and the reference temperature (To) when the reference charging time is recorded can be input.

[0123] The machine learning model can be trained in advance using comparison charging times measured for a test battery of the same type as the target battery.

[0124] Test batteries of the same type as the target battery can be prepared.

[0125] A machine learning model training device can measure the charging time (training comparison charging time) at each of multiple terminal voltage intervals while charging test batteries in constant current mode. For example, when multiple terminal voltage intervals are formed at intervals of 0.1V within the range of 3.0V to 4.2V, the machine learning model training device can measure the time required for the terminal voltage of the test batteries being charged at constant current to pass through the corresponding interval as the training comparison charging time. Additionally, the machine learning model training device can record the training comparison charging time measured for each terminal voltage interval.

[0126] The machine learning model training device can record training comparison charging times for each charge-discharge cycle while increasing the charge-discharge cycles of the test batteries. At this time, the machine learning model training device can also measure the SOH of the test batteries and record the SOH along with the training comparison charging times.

[0127] The machine learning model training device can measure and record training comparison charging times for each operating temperature. Test batteries can be separated into several groups, and each group can be distributed and placed in chambers with different temperatures. Additionally, the machine learning model training device can record training comparison charging times for each temperature.

[0128] In addition, the machine learning model training device can tune the internal parameters of the machine learning model by inputting the ratio value (training ratio value) of the reference charge time and the training comparison charge times into the machine learning model and comparing the result value with the pre-measured SOH.

[0129] A machine learning model training device can input training ratio values ​​and at least one terminal voltage value of the operating temperature and / or terminal voltage range into the machine learning model, and tune the parameters within the machine learning model by comparing the result with a pre-measured SOH.

[0130] FIG. 11 is a diagram of the fourth example configuration of a machine learning model according to the first embodiment.

[0131] Referring to FIG. 11, the machine learning model (1110) may be in the form of L (L is a natural number greater than or equal to 2) sub-machine learning models (1120a to 1120l) combined in the form of an ensemble.

[0132] The first sub-machine learning model (1120a) may be a model trained by training data of the first terminal voltage range—ratio value, SOH, operating temperature, etc., the second sub-machine learning model (1120b) may be a model trained by training data of the second terminal voltage range—ratio value, SOH, operating temperature, etc., and the L sub-machine learning model (1120l) may be a model trained by training data of the L terminal voltage range—ratio value, SOH, operating temperature, etc.

[0133] The sub-machine learning models (1120a to 1120l) can be combined in the form of an ensemble and then further trained with the training data of each terminal voltage range. At this time, weights can be assigned to the output values ​​of each sub-machine learning model according to the accuracy of each sub-machine learning model, and the accuracy of the final output value (SOH) can be improved by these weights.

[0134] The machine learning model (1110) may include N sub-machine learning models, and the data input to each sub-machine learning model may be ratio values ​​calculated at each of the N terminal voltage intervals mentioned above.

[0135] FIG. 12 is a diagram of the fifth example configuration of a machine learning model according to the first embodiment.

[0136] Referring to FIG. 12, the machine learning model (1210) may be composed of L (L is a natural number greater than or equal to 2) sub-machine learning models (1220a to 1220l).

[0137] The first sub-machine learning model (1220a) may be a model trained by training data of the first terminal voltage range—ratio value, SOH, operating temperature, etc., the second sub-machine learning model (1220b) may be a model trained by training data of the second terminal voltage range—ratio value, SOH, operating temperature, etc., and the L sub-machine learning model (1220l) may be a model trained by training data of the L terminal voltage range—ratio value, SOH, operating temperature, etc.

[0138] Sub-machine learning models (1120a to 1120l) can each output a SOH. The first sub-machine learning model (1220a) can output a first SOH, the second sub-machine learning model (1220b) can output a second SOH, and the L-sub-machine learning model (1220l) can output a first SOH.

[0139] Each sub-machine learning model (1120a to 1120l) may have a priority assigned to it. The estimation device can use the SOH calculated by the sub-machine learning model with the highest priority as the aging state value of the target battery. According to this method, there is an advantage that the aging state of the target battery can be estimated even if input data—e.g., ratio values—is not generated over the entire range and is generated only in a partial range.

[0140] FIG. 13 is a flowchart of a method for estimating battery aging state according to the first embodiment.

[0141] Referring to FIG. 13, the estimation device can store the charging time measured at each of a plurality of terminal voltage intervals for a reference aging state battery that is charged with constant current as the reference charging time (S1300).

[0142] The size of each terminal voltage interval may be the same. Alternatively, the sizes of at least two terminal voltage intervals may differ, in which case the reference charging time in at least two terminal voltage intervals may have similar sizes within a certain error range.

[0143] The estimation device can store the charging time for each of N (N is a natural number greater than or equal to 2) terminal voltage intervals as a comparison charging time in the constant current charging mode of the target battery, and calculate the ratio value of the reference charging time and the comparison charging time that was stored in advance for each of the N terminal voltage intervals (S1302).

[0144] In step S1302, the estimation device may set N intervals as terminal voltage intervals from the terminal voltage of the target battery confirmed at a given point in time. In another aspect, the estimation device may calculate N ratio values ​​in any terminal voltage range.

[0145] The estimation device can measure comparative charging times for M terminal voltage intervals greater than N. In this case, the estimation device can set N terminal voltage intervals such that predetermined terminal voltage intervals are included among the M terminal voltage intervals. Here, the charging time in one terminal voltage interval belonging to the N terminal voltage intervals among the M terminal voltage intervals may be longer than the charging time in another terminal voltage interval that does not belong to the N terminal voltage intervals.

[0146] The estimation device can estimate the aging state of the target battery by inputting the calculated N ratio values ​​into a pre-trained machine learning model (S1304).

[0147] When estimating the aging state, the estimation device can estimate the aging state of the target battery by inputting the difference between the operating temperature and the reference temperature of the target battery into the machine learning model.

[0148] When estimating the aging state, the estimation device can estimate the aging state of the target battery by additionally inputting at least one terminal voltage value corresponding to N terminal voltage intervals into the machine learning model.

[0149] The machine learning model may be in the form of an ensemble of N sub-machine learning models that take the ratio value of each terminal voltage range as input and the aging state value as output.

[0150] The machine learning model is composed of L (where L is a natural number greater than N) sub-machine learning models, each taking the ratio value of each terminal voltage range as input and the aging state value as output, and each sub-machine learning model has a set priority, and when the estimation device estimates the aging state, it can estimate the aging state of the target battery based on the output value of the sub-machine learning model with the highest priority among the output values ​​of the N sub-machine learning models corresponding to the N terminal voltage ranges among the L sub-machine learning models.

[0151] FIG. 14 is a flowchart of a method for estimating battery aging state according to a second embodiment.

[0152] Referring to FIG. 14, the estimation device (1400) may include a measurement circuit (1410), a storage circuit (1420), and a state estimation circuit (1430), etc.

[0153] The measurement circuit (1410) can record terminal voltage values ​​of the target battery while counting the charging time when the target battery is being charged in a constant current charging mode. The recorded content may be in the form of a time value and a voltage value paired together, for example, in the form of (00:00:00:000, 0V).

[0154] The measurement circuit (1410) can record time and terminal voltage in fixed time units—for example, in 1-second units. Or the measurement circuit (1410) can record time and terminal voltage in fixed voltage units—for example, in 0.1V units. Or the measurement circuit (1410) can record time and terminal voltage according to a predetermined standard. For example, the measurement circuit (1410) can record time and terminal voltage at a predetermined specific terminal voltage, and can record time and terminal voltage at a predetermined specific charging time.

[0155] The storage circuit (1420) can store terminal voltage time change rate data that records terminal voltage time change rates in multiple terminal voltage intervals according to aging state.

[0156] The terminal voltage time rate of change is a value representing the amount of change in terminal voltage for a constant charging time, and can be calculated by dividing the amount of change in terminal voltage (ΔV) by the charging time (Δt).

[0157] The storage circuit (1420) can store the terminal voltage time change rate for each terminal voltage interval for a plurality of terminal voltage intervals while increasing the number of charge / discharge cycles for a battery of the same type as the target battery. The data stored in this way can be called reference data.

[0158] FIG. 15 is an example drawing of a lookup table that stores the terminal voltage time change rate for each charge / discharge cycle in the second embodiment.

[0159] Referring to FIG. 15, the storage circuit can divide the terminal voltage into multiple intervals and store the terminal voltage time change rate (ΔV / Δt) in each terminal voltage interval.

[0160] The storage circuit can measure and store the terminal voltage time change rate for each charge / discharge cycle for test batteries used to generate reference data. For example, the storage circuit can store the terminal voltage time change rate for multiple terminal voltage intervals for a test battery with 1 charge / discharge cycle, and can store the terminal voltage time change rate for multiple terminal voltage intervals for a test battery with 364 charge / discharge cycles.

[0161] The storage circuit can measure and store SOH at each charge / discharge cycle. For example, the storage circuit can store the terminal voltage time change rate for a test battery with 1 charge / discharge cycle and also measure and store SOH. Additionally, the storage circuit can store the terminal voltage time change rate for a test battery with 364 charge / discharge cycles and also measure and store SOH.

[0162] Data can be generated for multiple test batteries. The storage circuit can statistically process the data generated for multiple test batteries to generate a single reference data. For example, the storage circuit can generate a single reference data by averaging the data generated for multiple test batteries.

[0163] The terminal voltage time change rate for each terminal voltage interval can be measured and stored for each operating temperature.

[0164] FIG. 16 is an example drawing of lookup tables stored by temperature in the second embodiment.

[0165] Referring to Fig. 16, the storage circuit can store lookup tables for each temperature.

[0166] Test batteries can be distributed and placed in chambers with different temperatures, and values ​​required for the aforementioned lookup table can be calculated while being charged and discharged under different temperature conditions.

[0167] In addition, the storage circuit can store the values ​​calculated during this process—for example, the terminal voltage time change rate by terminal voltage range, SOH, etc.—by classifying them according to temperature.

[0168] The estimation device can identify the data in a state most similar to the target battery from the reference data stored in the storage circuit and estimate the aging state of the target battery using the SOH value corresponding to that data.

[0169] Referring again to FIG. 14, the estimation device (1400) includes a measurement circuit (1410) and a storage circuit (1420), and may further include a state estimation circuit (1430).

[0170] The state estimation circuit (1430) can calculate the time rate of change of the terminal voltage in each of N (N is a natural number greater than or equal to 2) terminal voltage intervals using the charging time and terminal voltage values ​​recorded in the measurement circuit (1410).

[0171] And, the state estimation circuit (1430) can estimate the aging state of the target battery by comparing the calculated N terminal voltage time change rates with the terminal voltage time change rate data stored in the storage circuit (1420)—reference data described with reference to FIG. 15 and FIG. 16.

[0172] FIG. 17 is a diagram illustrating the process of finding the data most similar to the terminal voltage time change rate of the target battery in the second embodiment.

[0173] Referring to FIG. 17, the state estimation circuit can check the operating temperature of the target battery and first find the lookup table measured at the temperature closest to the operating temperature. In FIG. 17, the reference temperature is shown as being closest to the operating temperature of the target battery.

[0174] After finding a lookup table for operating temperature, the state estimation circuit can identify N terminal voltage ranges measured for the target battery among the multiple terminal voltage ranges stored in the lookup table. For example, if the lookup table stores terminal voltage time change rates for multiple terminal voltage ranges corresponding to the range of 3.1V to 4.2V for the test battery, and the N terminal voltage ranges measured for the target battery correspond to the range of 3.2V to 3.7V, the state estimation circuit can identify terminal voltage range data corresponding to the range of 3.2V to 3.7V in the lookup table.

[0175] When the portion corresponding to N terminal voltage intervals is defined as the search window, the state estimation circuit can search for data most similar to the time rate of change of the target battery's terminal voltage by moving the search window through the data sorted by charge / discharge cycles. Once the search is complete, the state estimation circuit can determine the SOH corresponding to the data as the aging state value of the target battery.

[0176] When forming a vector of the time rate of change of terminal voltage for N terminal voltage intervals, the state estimation circuit can determine the aging state value of the target battery by determining the similarity between the vectors stored in the lookup table and the vector corresponding to the target battery.

[0177] Similarity can be determined based on the Euclidean distance. The state estimation circuit can calculate Euclidean distance values ​​for each of the vectors corresponding to the target battery and the vectors stored in the lookup table. Then, the state estimation circuit can determine the vector corresponding to the shortest distance value among the calculated Euclidean distance values ​​as the vector with the highest similarity and determine the SOH corresponding to that vector as the aging state value of the target battery.

[0178] FIG. 18 is a flowchart of a method for estimating battery aging state according to a second embodiment.

[0179] Referring to FIG. 18, the estimation device can calculate the time rate of change of terminal voltage in each of N (N is a natural number greater than or equal to 2) terminal voltage intervals in the constant current charging mode of the target battery (S1800).

[0180] At least two of the N terminal voltage ranges may have different magnitudes. For example, the first terminal voltage range may be a range corresponding to 3.00V to 3.05V with a magnitude of 0.05V, and the second terminal voltage range may be a range corresponding to 3.05V to 3.15V with a magnitude of 0.1V.

[0181] The section with a large terminal voltage time change rate can be set to have a smaller section size than the section with a small terminal voltage time change rate.

[0182] The estimation device can estimate the aging state of the target battery by comparing the N terminal voltage time change rates calculated in the above step with the terminal voltage time change rate data stored for each aging state (S1802).

[0183] The terminal voltage time change rate data may include a lookup table that stores the terminal voltage time change rates in multiple terminal voltage intervals according to the aging state.

[0184] When estimating the aging state, the estimation device can find data in a lookup table that stores terminal voltage time change rates with high similarity to N terminal voltage time change rates, and determine the aging state value corresponding to the data as the aging state value of the target battery.

[0185] When estimating the aging state, the estimation device can calculate a Euclidean distance value for the N calculated terminal voltage time change rates and the terminal voltage time change rates stored in the lookup table, find the data with the smallest Euclidean distance value, and determine the aging state value corresponding to the data as the aging state value of the target battery.

[0186] The range of multiple terminal voltage intervals corresponding to the lookup table may be wider than the range of N terminal voltage intervals calculated for the target battery.

[0187] When estimating the aging state, the estimation device selects a lookup table corresponding to the temperature value of the target battery among the lookup tables included in the terminal voltage time change rate data, and can determine the aging state value of the target battery according to the selected lookup table.

[0188] One embodiment has the advantage of being able to effectively estimate the aging state of a battery even with partial charging, and enables online SOH estimation due to the simple calculation.

[0189] Terms such as "include," "compose," or "have" as described above, unless specifically stated otherwise, mean that the relevant component may be inherent; therefore, they should be interpreted as allowing for the inclusion of additional components rather than excluding them. All terms, including technical or scientific terms, have the same meaning as generally understood by those skilled in the art to which the present invention pertains, unless otherwise defined. Commonly used terms, such as those defined in advance, should be interpreted in accordance with their meaning in the context of the relevant technology and should not be interpreted in an ideal or overly formal sense unless explicitly defined in the present invention.

[0190] The foregoing description is merely an illustrative explanation of the technical concept of the present invention, and those skilled in the art to which the present invention pertains will be able to make various modifications and variations within the scope of the essential characteristics of the present invention. Accordingly, the embodiments disclosed in the present invention are intended to explain, not limit, the technical concept of the present invention, and the scope of the technical concept of the present invention is not limited by these embodiments. The scope of protection of the present invention shall be interpreted by the claims below, and all technical concepts within an equivalent scope shall be interpreted as being included within the scope of rights of the present invention.

Claims

Claim 1 A method for estimating the aging state of a battery, comprising: a step in which a battery aging state estimator stores a charging time measured at each of a plurality of terminal voltage intervals as a reference charging time for a battery in a reference aging state that is charged with constant current; a step in which the battery aging state estimator stores a charging time for each of N (N is a natural number greater than or equal to 2) terminal voltage intervals as a comparison charging time in a constant current charging mode of a battery, and calculates a ratio value between the reference charging time and the comparison charging time for each of the N terminal voltage intervals; and a step in which the battery aging state estimator inputs the N ratio values ​​into a pre-trained machine learning model to estimate the aging state of the battery, wherein when the comparison charging time is measured for M (M is a natural number greater than N) terminal voltage intervals, the N terminal voltage intervals are set such that predetermined terminal voltage intervals are included in the step of calculating the ratio value. Claim 2 A method for estimating the aging state of a battery according to claim 1, wherein, in the step of estimating the aging state, the difference between the operating temperature and the reference temperature of the battery is further input into the machine learning model to estimate the aging state of the battery. Claim 3 A method for estimating the aging state of a battery, wherein the size of each of the N terminal voltage intervals is the same in claim 1. Claim 4 A method for estimating the aging state of a battery according to claim 1, wherein, in the step of estimating the aging state, at least one terminal voltage value corresponding to the N terminal voltage intervals is further substituted into the machine learning model to estimate the aging state of the battery. Claim 5 A method for estimating the aging state of a battery according to claim 1, wherein, in the step of calculating the ratio value, N intervals are set from the terminal voltage of the battery confirmed at one point in time as the N terminal voltage intervals. Claim 6 delete Claim 7 A method for estimating the aging state of a battery according to claim 1, wherein the charging time in one terminal voltage section belonging to the N terminal voltage sections among the M terminal voltage sections is longer than the charging time in another terminal voltage section not belonging to the N terminal voltage sections. Claim 8 A method for estimating the aging state of a battery according to claim 1, wherein the machine learning model is a form in which N sub-machine learning models, each having a ratio value of each terminal voltage interval as input and an aging state value as output, are combined in the form of an ensemble. Claim 9 A step in which a battery aging state estimation device stores the charging time measured at each of a plurality of terminal voltage intervals as a reference charging time for a reference aging state battery being charged with constant current; a step in which, in a constant current charging mode of a battery, the battery aging state estimation device stores the charging time for each of N (N is a natural number greater than or equal to 2) terminal voltage intervals as a comparison charging time, and calculates the ratio value of the reference charging time and the comparison charging time for each of the N terminal voltage intervals; A battery aging state estimation device comprises a step of estimating the aging state of a battery by inputting N ratio values ​​into a pre-trained machine learning model, wherein the machine learning model is composed of L (where L is a natural number greater than N) sub-machine learning models that take ratio values ​​of each terminal voltage range as input and aging state values ​​as output, and each sub-machine learning model has a set priority, and in the step of estimating the aging state, the aging state of the battery is estimated according to the output value of the sub-machine learning model with the highest priority among the output values ​​of the N sub-machine learning models corresponding to the N terminal voltage ranges among the L sub-machine learning models. Claim 10 A method for estimating the aging state of a battery according to claim 9, wherein at least two of the N terminal voltage intervals have different sizes, and the reference charging time or the comparison charging time in the at least two terminal voltage intervals have similar sizes within a certain error range. Claim 11 A battery aging state estimation device comprising: a storage circuit that stores a charging time measured at each of a plurality of terminal voltage intervals as a reference charging time for a battery in a reference aging state that is charged with constant current; a calculation circuit that stores a charging time for each of N (N is a natural number greater than or equal to 2) terminal voltage intervals as a comparison charging time in a constant current charging mode of a battery, and calculates a ratio value between the reference charging time and the comparison charging time for each of the N terminal voltage intervals; and a state estimation circuit that estimates the aging state of the battery by inputting the N ratio values ​​into a pre-trained machine learning model, wherein when the comparison charging time is measured for M (M is a natural number greater than N) terminal voltage intervals, the calculation circuit sets the N terminal voltage intervals such that predetermined terminal voltage intervals are included. Claim 12 ◈Claim 12 was abandoned upon payment of the registration fee.◈ In claim 11, the state estimation circuit is a battery aging state estimation device that estimates the aging state of the battery by further inputting the difference between the operating temperature and the reference temperature of the battery into the machine learning model. Claim 13 ◈Claim 13 was abandoned upon payment of the registration fee.◈ In claim 11, the state estimation circuit is a battery aging state estimation device that estimates the aging state of the battery by further substituting at least one terminal voltage value corresponding to the N terminal voltage intervals into the machine learning model. Claim 14 ◈Claim 14 was abandoned upon payment of the registration fee.◈ In claim 11, the calculation circuit is a battery aging state estimation device that sets N intervals from the terminal voltage of the battery confirmed at one point in time as the N terminal voltage intervals. Claim 15 delete

Images