Learning method of open-circuit voltage estimation model of secondary battery, open-circuit voltage estimation method, and state estimation device

Abstract

The present application provides a learning method of an open-circuit voltage estimation model of a secondary battery, an open-circuit voltage estimation method, and a state estimation device. The open-circuit voltage of a secondary battery in operation is estimated with high accuracy. The method includes the following steps: measuring state variables including a terminal current and a terminal voltage of the secondary battery in operation at a prescribed time interval; preprocessing the state variables and calculating voltage estimation input data; and learning, by machine learning, a relationship between the voltage estimation input data and the open-circuit voltage of the secondary battery, in the step of calculating, a change rate of a voltage differential with respect to a current differential, i.e., a differential gradient, is calculated based on current differential time series data of the terminal current and voltage differential time series data of the terminal voltage, and voltage estimation input data including time series data of the terminal current, time series data of the terminal voltage, and time series data of the differential gradient is generated.

Description

Technical Field

[0001] This invention relates to a learning method for an open-circuit voltage estimation model for estimating the open-circuit voltage of a secondary battery in operation, an open-circuit voltage estimation method, and a state estimation method for a secondary battery using the learned open-circuit voltage estimation model. Background Technology

[0002] Rechargeable batteries, also known as secondary batteries, are widely used in electric vehicles, electric bicycles, and other mobile devices or buildings. When using these secondary batteries, it is important to understand their condition in order to determine the appropriate charging and replacement times. Here, the condition of a secondary battery refers to its SOC (State of Charge) and / or SOH (State of Health).

[0003] Typically, the charge-discharge characteristics of a secondary battery depend on its electrical characteristics, such as its SOC-OCV (Open Circuit Voltage) characteristics and / or the dependence of these characteristics on its SOH (State of Health). Therefore, from the viewpoint of accurate SOC estimation, the OCV of the secondary battery is desired, for example, as input to a state estimation model using a neural network. However, in secondary batteries, such as those connected to a vehicle's drive motor, which undergo frequent and repeated discharge and regeneration (charging) during driving, accurate measurement of the OCV is difficult.

[0004] Therefore, the conventional approach is to measure the internal resistance of the secondary battery, which is highly correlated with the OCV, instead of measuring the OCV of the secondary battery during operation, and use this measured internal resistance as input to the aforementioned state estimation model (see, for example, Patent Document 1). Typically, this internal resistance measurement is performed by inputting an AC signal for internal resistance measurement between the terminals of the secondary battery, overlapping with the charging and discharging current of the secondary battery.

[0005] However, the measurement of internal resistance based on the overlap of AC signals only indirectly represents the current OCV, and this measured value may not accurately represent the current OCV. Furthermore, the electrical characteristics of a secondary battery can vary depending on the manufacturer and / or model. Therefore, in the aforementioned prior art, where the measured values ​​of the secondary battery's voltage V, current I, and internal impedance Z are directly used as inputs to the neural network during learning, it is difficult to accurately estimate the SOC and other parameters of secondary batteries from various manufacturers and / or models with different electrical characteristics.

[0006] That is, when performing state estimation of secondary batteries, if the current OCV of various operating secondary batteries with different electrical characteristics can be estimated with high accuracy, the estimation accuracy can be improved when secondary batteries from various manufacturers and / or models are used as objects for state estimation.

[0007] Existing technical documents

[0008] Patent documents

[0009] Patent Document 1: Japanese Patent Application Publication No. 2003-249271 Summary of the Invention

[0010] The problem the invention aims to solve

[0011] The present invention was made in view of the above circumstances, and its object is to estimate the open-circuit voltage (OCV) of secondary batteries in various operations with different electrical characteristics with high accuracy.

[0012] means for solving problems

[0013] One aspect of the present invention is a learning method for an open-circuit voltage estimation model of a secondary battery. This method is a machine learning-based approach for estimating the open-circuit voltage of a working secondary battery connected to a load or charger. The learning method includes the following steps: measuring state variables, including terminal current and terminal voltage, of the working secondary battery at predetermined time intervals; preprocessing the state variables and calculating voltage estimation input data; and using machine learning to enable the open-circuit voltage estimation model to learn the voltage estimation input data and the open-circuit voltage of the working secondary battery. Regarding the relationship between voltages, in the calculation step, based on the time series data of the terminal current and the time series data of the terminal voltage, the difference between the terminal current (i.e., current difference) and the difference between the terminal voltage (i.e., voltage difference) are calculated. Based on the time series data of the current difference and the time series data of the voltage difference, the rate of change of the voltage difference relative to the current difference (i.e., differential gradient) is calculated from the period from the past, which traces back to the first predetermined time, to the present. This generates the voltage estimation input data, which includes the time series data of the terminal current, the time series data of the terminal voltage, and the time series data of the differential gradient.

[0014] According to another aspect of the invention, the differential gradient is calculated using the least squares method.

[0015] According to another aspect of the present invention, the current difference and the voltage difference are respectively the fourth-order difference of the time series data of the terminal current and the fourth-order difference of the time series data of the terminal voltage.

[0016] According to another aspect of the present invention, the open-circuit voltage estimation model is composed of a recurrent neural network (RNN).

[0017] According to another aspect of the present invention, the intermediate layer of the RNN constituting the open-circuit voltage estimation model is composed of a Long Short-Term Memory (LSTM) or a Gated Recursive Unit (GRU).

[0018] According to another aspect of the present invention, the open-circuit voltage estimation model is composed of a one-dimensional convolutional neural network (CNN).

[0019] According to another aspect of the invention, the open-circuit voltage estimation model is generated by learning from time-series data of state variables, including terminal current and terminal voltage, of multiple secondary batteries with different electrical characteristics connected to a load or charger.

[0020] Another aspect of the present invention is a method for estimating the open-circuit voltage of a secondary battery, wherein the method includes the following steps: measuring state variables of the secondary battery in operation, including terminal current and terminal voltage, at predetermined time intervals; preprocessing the state variables and calculating voltage estimation input data; and estimating the open-circuit voltage of the secondary battery in operation using the open-circuit voltage estimation model learned through any of the above-mentioned learning methods, based on the voltage estimation input data. In the calculation step, based on the time series data of the terminal current and the time series data of the terminal voltage, the difference between the terminal current (i.e., current difference) and the difference between the terminal voltage (i.e., voltage difference) are calculated. Based on the time series data of the current difference and the time series data of the voltage difference, the rate of change of the voltage difference relative to the current difference (i.e., differential gradient) is calculated from the past, tracing back to the present, to the present, to generate the voltage estimation input data including the time series data of the terminal current, the time series data of the terminal voltage, and the time series data of the differential gradient.

[0021] Another aspect of the present invention is a state estimation device for a secondary battery, which estimates the state of a secondary battery in operation. The state estimation device comprises: a state observation unit that measures state variables including terminal current and terminal voltage of the secondary battery in operation at predetermined time intervals; a preprocessing unit that preprocesses the state variables measured by the state observation unit and calculates input data; and a state estimation unit that estimates the current charging rate and / or degradation degree of the secondary battery in operation based on the input data. The state estimation unit uses an open-circuit voltage estimation model learned through any of the above-described learning methods to estimate the current open-circuit voltage of the secondary battery in operation, and uses the estimated open-circuit voltage to estimate the current charging rate and / or degradation degree of the secondary battery in operation.

[0022] The effects of the invention

[0023] According to the present invention, the open-circuit voltage of a working secondary battery with various electrical characteristics from different manufacturers or models can be estimated with high accuracy. Attached Figure Description

[0024] Figure 1 This is a flowchart illustrating the steps of the learning method for the open-circuit voltage estimation model according to the first embodiment of the present invention.

[0025] Figure 2 Is to show execution Figure 1 A diagram showing the structure of a machine learning device for the learning method of the open-circuit voltage estimation model.

[0026] Figure 3 It is shown Figure 1 The flowchart shows a detailed process for processing the input data for voltage estimation in the learning method of the open-circuit voltage estimation model.

[0027] Figure 4 It is used for Figure 3 The diagram illustrates the calculation of the current difference in the process shown.

[0028] Figure 5 It is used for Figure 3 The diagram illustrates the calculation of the voltage differential in the process shown.

[0029] Figure 6 It is used for Figure 3 The diagram illustrates the calculation of the differential gradient in the processing shown.

[0030] Figure 7 It is shown Figure 2 A diagram illustrating an example of the structure of an open-circuit voltage estimation model generated by the model learning unit of the machine learning device shown.

[0031] Figure 8 This is a diagram illustrating an example of open-circuit voltage estimation for a secondary battery using a learned open-circuit voltage estimation model.

[0032] Figure 9 This is a flowchart illustrating the steps of the open-circuit voltage estimation method according to the second embodiment of the present invention.

[0033] Figure 10 Is to show execution Figure 9 A diagram showing the structure of the state estimation device for the open-circuit voltage estimation method.

[0034] Figure 11 yes Figure 10 The diagram shows the functional block diagram of the processing unit of the state estimation device.

[0035] Label Explanation

[0036] 100…Machine learning device, 102, 404…Secondary battery, 104…Charger, 106…Load, 108…Switch, 110, 406…Characteristic measuring device, 112…Learning management device, 120, 420, 440…Processing device, 122, 428, 448…Storage device, 124, 430…Open circuit voltage estimation model, 130…State variable measuring unit, 132…Input data generation unit, 134…Model learning unit, 200, 202, 204…Approximate straight line, 300, 310…Input layer, 302, 312…Intermediate layer, 304, 314…Output layer, 400…State variable measurement unit, 120…Approximate straight line ... 402…vehicle, 408…power controller, 410…rotary motor, 412…external charging device, 414…driving control device, 422…state observation unit, 424…preprocessing unit, 426…state estimation unit, 442…motor control unit, 444…charging control unit, 446…notification control unit, 450…display device, 452…accelerator pedal sensor, 454…brake pedal sensor, 456…vehicle speed sensor, 500, 502, 504, 506, 508, 510, 512, 514, 516, 518…processing, 600, 602, 604, 606…lines. Detailed Implementation

[0037] The inventors of this invention have discovered a correlation between, at least among the same type of secondary batteries (e.g., secondary batteries of the same type, "lithium-ion batteries"), the higher-order variation patterns of the terminal current and terminal voltage of the secondary battery and the internal states (OCV, SOC, and / or SOH) of the secondary battery. Furthermore, the inventors have realized that by using the rate of change (the difference gradient described later) of the time-series data of the terminal voltage relative to the time-series data of the terminal current as a parameter representing the higher-order variation patterns of the terminal current and terminal voltage of the secondary battery, and by using this as input to a model (e.g., a neural network), a model capable of accurately estimating the state of secondary batteries with various electrical characteristics from different manufacturers or models can be generated. This invention is based on this insightful observation.

[0038] Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings.

[0039] [First Implementation]

[0040] Figure 1 This diagram illustrates the steps of a learning method for an open-circuit voltage estimation model of a secondary battery according to a first embodiment of the present invention. The learning method includes the following steps: measuring state variables, including terminal current and terminal voltage, of a working secondary battery connected to a load or charger at predetermined time intervals (S100); and preprocessing the measured state variables and calculating voltage estimation input data (S102). Furthermore, the learning method includes a step of using machine learning to enable the open-circuit voltage estimation model to learn the relationship between the calculated voltage estimation input data and the open-circuit voltage of the working secondary battery (S104).

[0041] Figure 2 Is to show execution Figure 1 The diagram illustrates an example of the structure of a learning management device and a machine learning device for a learning method of the open-circuit voltage estimation model. The open-circuit voltage estimation model is, for example, constructed from a neural network. The learning management device 112 controls the operation of the secondary battery 102 during the aforementioned machine learning process and calculates the measured value of the open-circuit voltage as teaching data, providing it to the machine learning device 100.

[0042] The secondary battery 102 is charged by the charger 104 and discharged by supplying power to the load 106. The charger 104 is, for example, a DC power supply, and the load 106 is, for example, a motor. The selection of whether to charge the secondary battery 102 from the charger 104 or to discharge the secondary battery 102 to the load 106 is made by a switch 108. A characteristic measuring device 110 is inserted between the switch 108 and the secondary battery 102.

[0043] The characteristic measuring device 110 measures the current value of a specified state variable of the secondary battery 102. The specified state variable may include the terminal voltage Vte, terminal current Ite, internal impedance Z, and the temperature T (°C) of the casing surface of the secondary battery 102. Here, the internal impedance Z can be measured, for example, by inputting an alternating current as a measurement signal into the secondary battery 102, according to the prior art.

[0044] The terminal current Ite of the secondary battery 102 is positive when the secondary battery 102 is discharging and negative when it is charging.

[0045] [1. Learning Management Device]

[0046] The learning management device 112 controls the charging and discharging of the secondary battery 102, and generates teaching data for learning the open-circuit voltage estimation model and outputs it to the machine learning device 100. The learning management device 112 is, for example, a computer, which starts operating according to instructions from the operator, giving instructions to start and stop the power output of the charger 104, and instructions to switch the toggle switch 108.

[0047] The learning management device 112 obtains the terminal current Ite, terminal voltage Vte, and internal impedance Z of the secondary battery 102 during charging and discharging from the characteristic measuring device 110 at predetermined time intervals.

[0048] The learning management device 112 calculates the open-circuit voltage Voc of the secondary battery 102 based on the terminal current Ite, terminal voltage Vte, and internal impedance Z obtained above, and generates time-series data of the open-circuit voltage Voc. This time-series data of the open-circuit voltage Voc is used as teaching data for the learning of the open-circuit voltage estimation model executed by the machine learning device 100 described later.

[0049] [2. Machine Learning Device]

[0050] Machine learning device 100 execution Figure 1The learning method for the open-circuit voltage estimation model is shown. The machine learning device 100 includes a processing unit 120 and a storage unit 122. The storage unit 122 is configured, for example, a volatile and / or non-volatile semiconductor memory and / or a hard disk device. The storage unit 122 stores the open-circuit voltage estimation model 124 generated by the model learning unit 134 described later.

[0051] The processing device 120 is, for example, a computer equipped with a processor such as a CPU (Central Processing Unit). The processing device 120 may also have a structure including ROM (Read Only Memory) for writing programs and RAM (Random Access Memory) for temporarily storing data. Furthermore, the processing device 120 includes a state variable measurement unit 130, an input data generation unit 132, and a model learning unit 134 as functional elements or units.

[0052] These functional elements of the processing device 120 are implemented, for example, by executing a program by the processing device 120, which is a computer. Furthermore, the aforementioned computer program can be pre-stored in any computer-readable storage medium. Alternatively, all or part of the aforementioned functional elements of the processing device 120 can be constructed from hardware comprising one or more electronic circuit components.

[0053] [2.1 Functions of the State Variable Measurement Unit]

[0054] State variable measurement unit 130 executes Figure 1 The step S100 is shown. That is, the state variable measuring unit 130 obtains state variables, including terminal current Ite and terminal voltage Vte, of the secondary battery 102 connected to the load 106 or charger 104 from the characteristic measuring unit 110 at predetermined time intervals. Thus, the state variable measuring unit 130 measures the state variables at predetermined time intervals. The state variable measuring unit 130 can also measure the temperature T of the secondary battery 102 as a state variable at the aforementioned predetermined time intervals.

[0055] [2.2 Functions of the Input Data Generation Unit]

[0056] Input data generation unit 132 executes Figure 1 The step S102 shown is as follows: The input data generation unit 132 preprocesses the state variables measured by the state variable measurement unit 130 and calculates the voltage estimation input data.

[0057] Figure 3 It is shown Figure 1The flowchart details the processing in step S102 of calculating the voltage estimation input data. In step S102, the input data generation unit 132 first calculates the difference between the terminal current Ite (i.e., the current difference δIte) and the difference between the terminal voltage Vte (i.e., the voltage difference δVte) based on the time series data of the terminal current Ite and the time series data of the terminal voltage Vte (S200). Next, the input data generation unit 132 calculates the rate of change of the voltage difference δVte relative to the current difference δIte, i.e., the differential gradient Sdiff, from the time elapsed to the present (a predetermined time T1 has been traced back to the present) (S202).

[0058] Then, the input data generation unit 132 generates state estimation input data containing the following three time series data from the period from the past, which traces back to the present time T2, to the present time (S204), and ends the processing.

[0059] Time series data of terminal current Ite

[0060] Time series data of terminal voltage Vte

[0061] Time series data of differential gradient Sdiff

[0062] The following explains the specific calculation methods for current difference δIte, voltage difference δVte, and differential gradient Sdiff.

[0063] [2.2.2.1 Calculation of current difference δIte]

[0064] Figure 4 This is a diagram used to illustrate the calculation of the current difference δIte. Figure 4 In the table shown, the leftmost column is designated as column 1, and the columns to the right are designated as column 2, column 3, ... column 6. Figure 4 The first column of the table indicates the time when the state variable measuring unit 130 repeatedly acquires the terminal current Ite according to the time interval dt, or the index (number) of that time. The second column is the time series data of the terminal current Ite, indicating the terminal current Ite acquired at each time.

[0065] Columns 3, 4, 5, and 6 represent the first-order difference Δ of the terminal current Ite calculated based on the terminal current Ite in column 2. 1 Ite, second-order difference Δ 2 Ite, third-order difference Δ 3 Ite, and the fourth-order difference Δ 4 Ite.

[0066] Current time t n h-th order difference Δ hIte(t n (h = 1, 2, ... 4) is calculated using the following formula.

[0067] Δ h Ite(t n )=Δ h-1 Ite(t n )-Δ h-1 Ite(t n-1 )

[0068] Here, h = 1, 2, 3, 4. Furthermore, let Δ... 0 Ite(t n ) = Ite(t n ).

[0069] That is, time t n First-order difference Δ 1 Ite(t n ) is from time t n Terminal current Ite(t) n Subtract time t n-1 Terminal current Ite(t) n-1 ) and calculated. Furthermore, time t n The second-order difference Δ 2 Ite(t n ) is from time t n First-order difference Δ 1 Ite(t n Subtract time t n-1 First-order difference Δ 1 Ite(t n-1 ) and calculated.

[0070] Similarly, at time t n The third difference Δ 3 Ite(t n ) is from time t n The second-order difference Δ 2 Ite(t n Subtract time t n-1 The second-order difference Δ 2 Ite(t n-1 And the calculation is based on time t. n The fourth-order difference Δ 4 Ite(t n ) is from time t n The third difference Δ 3 Ite(t n Subtract time t n-1 The third difference Δ 3 Ite(t n-1 ) and calculated.

[0071] In this embodiment, the input data generation unit 132 generates the fourth-order differential Δ of the terminal current Ite at each time. 4 Ite is set as the current difference δIte. That is,

[0072] δIte(t)=Δ 4 Ite(t)t=t n t n-1 、···.

[0073] [2.2.2.2 Calculation of voltage difference δVte]

[0074] The input data generation unit 132 calculates the voltage difference δVte of the terminal voltage Vte in the same way as the current difference described above. Figure 5 This is a diagram illustrating the calculation steps for the voltage difference δVte. In Figure 5 In the table shown, the leftmost column is designated as column 1, and the columns to the right are designated as column 2, column 3, ... column 6. Figure 5 The first column of the table indicates the time when the state variable measuring unit 130 repeatedly acquires the terminal voltage Vte according to the time interval dt, or the index (number) of that time. The second column is the time series data of the terminal voltage Vte, indicating the terminal voltage Vte acquired at each time.

[0075] Columns 3, 4, 5, and 6 represent the first-order difference Δ of the terminal voltage Vte calculated based on the terminal voltage Vte in column 2. 1 Vte, second-order difference Δ 2 Vte, third-order difference Δ 3 Vte, and the fourth-order difference Δ 4 Vte.

[0076] Current time t n h-th order difference Δ h Vte(t n (h = 1, 2, ... 4) is calculated using the following formula.

[0077] Δ h Vte(t n )=Δ h-1 Vte(t n )-Δ h-1 Vte(t n-1 )

[0078] Here, h = 1, 2, 3, 4. Furthermore, let Δ... 0 Vte(t n ) = Vte(t n ).

[0079] That is, time t n First-order difference Δ 1Vte(t n ) is from time t n Terminal voltage Vte(t) n Subtract time t n-1 Terminal voltage Vte(t) n-1 ) and calculated. Furthermore, time t n The second-order difference Δ 2 Vte(t n ) is from time t n First-order difference Δ 1 Vte(t n Subtract time t n-1 First-order difference Δ 1 Vte(t n-1 ) and calculated.

[0080] Similarly, at time t n The third difference Δ 3 Vte(t n ) is from time t n The second-order difference Δ 2 Vte(t n Subtract time t n-1 The second-order difference Δ 2 Vte(t n-1 And the calculation is based on time t. n The fourth-order difference Δ 4 Vte(t n ) is from time t n The third difference Δ 3 Vte(t n Subtract time t n-1 The third difference Δ 3 Vte(t n-1 ) and calculated.

[0081] In this embodiment, the input data generation unit 132 generates the fourth-order differential Δ of the terminal voltage Vte at each time step. 4 Vte is set as the voltage difference δVte. That is,

[0082] δVte(t)=Δ 4 Vte(t)t=t n t n-1 、···.

[0083] [2.2.2.3 Calculation of the difference gradient Sdiff]

[0084] The differential gradient Sdiff is the rate of change of the voltage differential δVte relative to the current differential δIte over a period from the present, tracing back a specified time T1 to the present. Specifically, as... Figure 4 and Figure 5As shown, the input data generation unit 132 extracts data from past time t. m up to the current time t n During the period, there are k1 (k1 = n - m + 1) current differences δIte and voltage differences δVte, where from the past time t m up to the current time t n The period is equivalent to the period from the past, which is traced back to a specified time T1, up to the present. Moreover, the input data generation unit 132 calculates the differential gradient Sdiff, which is the rate of change of the voltage differential δVte relative to the current differential δIte, using the least squares method based on the dataset (δIte, δVte) of each moment consisting of the extracted δIte and δVte.

[0085] More specifically, such as Figure 6 As shown, the slope of the approximate straight line (regression line) 200 of the plotted points (black circles inside the dashed ellipse) of the aforementioned k1 datasets (δIte, δVte) on a two-dimensional plane with the current difference δIte as the horizontal axis and the voltage difference δVte as the vertical axis is equivalent to the difference gradient Sdiff. That is, when the approximate straight line 200 is given by δVte = a1 × δIte + b1, the slope a1 of the approximate straight line 200 is equivalent to the difference gradient Sdiff. Here, the approximate straight line is calculated, for example, by the least squares method.

[0086] [2.2.2.5 Voltage Estimation Input Data]

[0087] As described above, the voltage estimation input data consists of the terminal current Ite, the terminal voltage Vte, and the differential gradient Sdiff, representing a time series data spanning from the present time, tracing back a specified time T2 to the present. Let the current time be t... n And let t be the moment that has been traced back to the present time T2. r At that time, the voltage estimation input data is represented by the following formula.

[0088] [Formula 1]

[0089]

[0090] in,

[0091]

[0092]

[0093]

[0094] Here is the time series data of the terminal current Ite. V Ite(t nTime series data of terminal voltage Vte V Vte(t n and time series data of differential gradient Sdiff v Sdiff(t n () are the terminal current Ite, terminal voltage Vte, and differential gradient Sdiff from time t. r At time t n The n-r+1 values ​​are used as the first-order tensor of the element. Therefore, the voltage estimation input data V x1(t n ) is a second-order tensor.

[0095] [2.3 Functions of the Model Learning Unit]

[0096] Model Learning Department 134 Execution Figure 1 In step S104 of the open-circuit voltage estimation model learning method shown, an open-circuit voltage estimation model 124 is generated through machine learning. Specifically, the model learning unit 134 uses the voltage estimation input data generated by the input data generation unit 132 to learn the open-circuit voltage estimation model 124 through machine learning. At this time, the model learning unit 134, for example, obtains time-series data of the open-circuit voltage Voc of the secondary battery 102 from the learning management device 112, and uses the obtained time-series data of the open-circuit voltage Voc as teaching data to perform the aforementioned machine learning.

[0097] Figure 7 This is a diagram illustrating the structure of the open-circuit voltage estimation model 124 generated by the model learning unit 134. The open-circuit voltage estimation model 124 is composed of a neural network, having an input layer 300, an intermediate layer 302, and an output layer 304. The open-circuit voltage estimation model 124 is, for example, an RNN (Recurrent Neural Network).

[0098] The input layer 300 accepts voltage estimation input data as a second-order tensor as shown in equation (1) above. The intermediate layer 302 in this embodiment includes a multi-layered LSTM (Long Short Term Memory). However, the intermediate layer 302 is not limited to LSTM. For example, the intermediate layer 302 may also be constructed from a GRU (Gated Recurrent Unit).

[0099] Output layer 304 outputs the time t of secondary battery 102. n The estimated value of the open-circuit voltage Voc is used as the output y1(t) n That is, the output is y1(t). n Voc(t) is the open-circuit voltage as a scalar.n ).

[0100] [3. Secondary batteries for model learning]

[0101] As the secondary battery 102 used for model learning, it is desirable to use a variety of secondary batteries with different electrical characteristics due to different manufacturers or models. This allows for the generation of an open-circuit voltage estimation model 124 with minimal variation in estimation accuracy across different manufacturers or models. For example, when learning the open-circuit voltage estimation model 124, it is desirable to use multiple secondary batteries with different electrical characteristics such as SOC-OCV characteristics, SOC-internal impedance characteristics, and / or their SOH dependence.

[0102] [4. The Actions of a Secondary Battery in Model Learning]

[0103] Regarding the behavior (charge / discharge story) of the secondary battery in the model learning, it is not merely expected that it will monotonously discharge or charge between a fully charged state (SOC = 100%) and a fully discharged state (SOC = 0%), but also that it will randomly charge and discharge and / or alternately charge and discharge according to a prescribed benchmark. This prescribed benchmark can be a benchmark corresponding to the intended use of the secondary battery being estimated. For example, when assuming a vehicle-use secondary battery as the estimation object, a benchmark adjusted according to the typical charge / discharge cycle of vehicles in various traffic scenarios such as urban areas, mountainous areas, rural areas, and highways can be used.

[0104] [5. Collection of learning data]

[0105] In this embodiment, the machine learning device 100 obtains the state variables (Ite, Vte) of the secondary battery 102, which form the basis of the learning data for the open-circuit voltage estimation model 124, from the characteristic measurer 110, as well as the time series data of OCV, which serves as teaching data. These data are then calculated by the learning management device 112 and immediately used for learning the open-circuit voltage estimation model 124. However, these state variables and the time series data of the teaching data do not necessarily have to be used for learning immediately.

[0106] The learning management device 112 can also pre-activate the secondary battery to acquire and store time-series data of state variables and teaching data. The machine learning device 100 can also acquire the time-series data of the aforementioned state variables and teaching data stored by the learning management device 112 from the learning management device 112 to learn the open-circuit voltage estimation model 124.

[0107] Furthermore, time series data of state variables and teaching data can also be generated by simulating the charging and discharging characteristics obtained from the design data such as the equivalent circuit of the secondary battery 102, as long as the error with the actual data is within a practically acceptable range.

[0108] [6. An example of open-circuit voltage estimation using an open-circuit voltage estimation model]

[0109] Next, an example of estimating the open-circuit voltage of a secondary battery using the open-circuit voltage estimation model learned through the learning method of this embodiment will be described. Figure 8 This is a diagram illustrating an example of open-circuit voltage estimation for a secondary battery using a learned open-circuit voltage estimation model.

[0110] The training data for the open-circuit voltage estimation model was generated by simulating the charge-discharge characteristics of dozens of sample secondary batteries for vehicles with different electrical characteristics using a computer. Specifically, for each of the dozens of sample secondary batteries with different electrical characteristics (SOC-OCV, internal impedance, and capacity (SOH), the computer simulation calculated the terminal current Ite, terminal voltage Vte, and open-circuit voltage at specified time intervals dt under the condition of charging and discharging according to a specified charge-discharge process.

[0111] The above charging and discharging process not only monotonically discharges or charges the sample secondary battery between a fully charged state (SOC=100%) and a fully discharged state (SOC=0%), but is also adjusted according to the typical charging and discharging cycle of vehicles in various traffic scenarios such as urban areas, mountainous areas, rural areas, and highways.

[0112] The sample secondary battery is a lithium-ion battery. Furthermore, the measurement interval dt for the state variable is 100 ms, and the specified times T1 and T2 for calculating the voltage estimation input data of the above open-circuit voltage estimation model are both 5 seconds. However, these times are just examples; the specified times T1 and T2 can also be set to different values.

[0113] Figure 8 The diagram shows the estimation results and simulated values ​​of the open-circuit voltage using the learned open-circuit voltage estimation model during the period when an arbitrary secondary battery selected from the above sample secondary batteries is discharged from a fully charged state to a fully discharged state.

[0114] exist Figure 8In the diagram, the horizontal axis represents the elapsed time since the start of discharge when the secondary battery begins to discharge from a fully charged state, and the vertical axis represents the open-circuit voltage (OCV) of the target secondary battery (unit: V). The voltage estimation input data provided to the open-circuit voltage estimation model in the open-circuit voltage estimation is calculated based on Ite and Vte, obtained through simulation of the target secondary battery's discharge at specified time intervals dt.

[0115] exist Figure 8 In the diagram, line 600, formed by the set of gray dots, represents the open-circuit voltage value calculated through simulation based on the charge-discharge characteristics of the target secondary battery. Furthermore, line 602, formed by the set of black dots, represents the estimated open-circuit voltage value obtained through an open-circuit voltage estimation model.

[0116] according to Figure 8 A comparison of lines 600 and 602 shows that the open-circuit voltage estimation model learned using the learning method described in this embodiment accurately estimates the open-circuit voltage of the target secondary battery. In particular, although the open-circuit voltage estimation model used for this estimation is generated using learning data from dozens of sample secondary batteries with different electrical characteristics, the open-circuit voltage estimates obtained from this model do not diverge but converge to a single line (line 602), accurately estimating the open-circuit voltage of a specific target secondary battery. Therefore, it can be seen that the open-circuit voltage estimation model learned using the learning method of this embodiment uses multiple secondary batteries with different electrical characteristics for learning, thereby enabling high-precision estimation of the open-circuit voltage of various operating secondary batteries from different manufacturers and models.

[0117] [Second Implementation]

[0118] Next, the second embodiment of the present invention will be described. Figure 9 This diagram illustrates the steps of an open-circuit voltage estimation method for a secondary battery according to an embodiment of the present invention. The open-circuit voltage estimation method includes the following steps: measuring state variables, including terminal current and terminal voltage, of a working secondary battery connected to a load or charger at predetermined time intervals (S300); and preprocessing the measured state variables and calculating voltage estimation input data (S302). Furthermore, the open-circuit voltage estimation method includes estimating the open-circuit voltage of the working secondary battery using an open-circuit voltage estimation model learned through the learning method of the first embodiment described above, based on the voltage estimation input data (S304).

[0119] Figure 9 The open-circuit voltage estimation method shown is, for example, in Figure 10This is performed in the state estimation device 400 shown. This state estimation device 400, for example, is mounted on a vehicle 402, which is an electric vehicle, and estimates the state of a secondary battery 404, which is the on-board battery of the vehicle 402, during operation. The secondary battery 404 is connected to a rotary motor 410 via a characteristic measuring device 406 and a power-on controller 408.

[0120] The rotary motor 410 functions as a motor that drives the wheels of the vehicle 402 by being powered by the discharge from the secondary battery 404, and also functions as a generator that generates electricity by generating electricity through the rotational force transmitted from the wheels to charge the secondary battery 404.

[0121] Characteristic measuring device 406 measures the current values ​​of state variables of secondary battery 404, including terminal current Ite and terminal voltage Vte. Power supply controller 408, under the control of driving control device 414 mounted on vehicle 402, controls the power supply from secondary battery 404 to rotary motor 410 and from rotary motor 410 to secondary battery 404. Furthermore, when an external charging device 412 located outside vehicle 402 is connected to vehicle 402, power supply controller 408, under the control of driving control device 414, controls the power supply from external charging device 412 to secondary battery 404. External charging device 412 is, for example, a charger at a charging station. Additionally, when another generator driven by an internal combustion engine is mounted on vehicle 402, power supply controller 408 can also control the power supply from that generator to secondary battery.

[0122] The driving control device 414 obtains estimated values ​​of the current SOC and SOH, representing the state of the secondary battery 404, from the state estimation device 400, controls the operation of the rotary motor 410 based on the obtained SOC and SOH, and notifies the user.

[0123] Specifically, the driving control device 414 includes a processing device 440 and a storage device 448. The storage device 448 is, for example, a semiconductor memory, which stores data required for processing by the processing device 440.

[0124] The processing device 440 is, for example, a computer equipped with a processor such as a CPU. The processing device 440 may also have a structure with a ROM for writing programs, RAM for temporarily storing data, etc. Moreover, the processing device 440 includes a motor control unit 442, a charging control unit 444, and a notification control unit 446 as functional elements or functional units.

[0125] These functional elements of the processing device 440 are implemented, for example, by executing a program by the processing device 440, which is a computer. Furthermore, the aforementioned computer program can be pre-stored in any computer-readable storage medium. Alternatively, all or part of the aforementioned functional elements of the processing device 440 can be constructed from hardware comprising one or more electronic circuit components.

[0126] The motor control unit 442 detects the amount of pressure applied to the accelerator pedal (not shown) of the vehicle 402 based on the accelerator pedal sensor 452. When the accelerator pedal is pressed, the driving control unit 414 instructs the power supply controller 408 to supply power from the secondary battery 404 to the rotary motor 410, causing the rotary motor 410 to operate as a motor and propel the vehicle 402. Furthermore, the driving control unit 414 controls the rotational speed of the rotary motor 410 via the power supply controller 408, ensuring that the speed of the vehicle 402 obtained from the vehicle speed sensor 456 corresponds to the amount of pressure applied to the accelerator pedal.

[0127] At this time, the motor control unit 442, based on the current SOC estimate obtained from the state estimation device 400, limits, for example, the upper limit (maximum current) of the current supplied from the secondary battery 404 to the rotary motor 410 during vehicle 402 acceleration or constant speed driving. That is, the motor control unit limits the discharge of the secondary battery 404, for example, to limit the torque generated by the rotary motor 410, and determines the maximum current in such a way that the fuel consumption (e.g., driving distance per 1 kWh) determined according to the characteristics of the secondary battery 404 and the rotary motor 410 is not less than a predetermined value.

[0128] The charging control unit 444 determines whether the brake pedal (not shown) of the vehicle 402 is depressed by the brake pedal sensor 454. Then, when the brake pedal is depressed, the charging control unit 444 instructs the motor control unit 442 to stop the power supply from the secondary battery 404 to the rotary motor 410. Then, the charging control unit 444 instructs the power supply controller 408 to supply power from the rotary motor 410 to the secondary battery 404, so that the rotary motor 410 operates as a generator to charge the secondary battery 404 (so-called regenerative braking action).

[0129] In addition, when the external charging device 412 is connected to the vehicle 402, the charging control unit 444 controls the amount of power supplied from the external charging device 412 to the secondary battery 404 via the power controller 408.

[0130] The notification control unit 446 displays a predetermined value on the display device 450 based on the current SOC and SOH estimates obtained from the state estimation device 400. For example, the notification control unit 446 only displays the obtained current SOC and SOH estimates on the display device 450. Furthermore, for example, when the SOC estimate is lower than a predetermined value, the notification control unit 446 displays a message on the display device 450 suggesting that the driver of vehicle 402 charge at a charging station. Alternatively, for example, when the SOH estimate is lower than a predetermined value, the notification control unit 446 displays a message on the display device 450 suggesting that the driver of vehicle 402 replace the secondary battery 404.

[0131] State estimation device 400 executes Figure 9 The open-circuit voltage estimation method shown estimates the open-circuit voltage of the secondary battery 404 during operation. Then, based on the estimated open-circuit voltage, the state estimation device 400 estimates the SOC and SOH of the secondary battery 404 and outputs the current SOC estimate and SOH estimate to the driving control device 414.

[0132] Specifically, the state estimation device 400 includes a processing device 420 and a storage device 428. The storage device 428 is composed of a non-volatile and a volatile semiconductor memory. In the storage device 428, an open-circuit voltage estimation model 124, which has been learned by the learning method of the first embodiment, is stored in advance as an open-circuit voltage estimation model 430.

[0133] The processing device 420 is, for example, a computer equipped with a processor such as a CPU. The processing device 420 may also have a structure with a ROM for writing programs, RAM for temporarily storing data, etc. Moreover, the processing device 420 includes a state observation unit 422, a preprocessing unit 424, and a state estimation unit 426 as functional elements or functional units.

[0134] These functional elements of the processing device 420 are implemented, for example, by executing a program by the processing device 420, which is a computer. Furthermore, the aforementioned computer program can be pre-stored in any computer-readable storage medium. Alternatively, all or part of the aforementioned functional elements of the processing device 420 can be constructed from hardware comprising one or more electronic circuit components.

[0135] Figure 11 A functional block diagram of a processing apparatus 420, comprising a state observation unit 422, a preprocessing unit 424, and a state estimation unit 426, is shown. Figure 11 In the diagram, the dashed rectangles represent the processing in the preprocessing unit 424.

[0136] Condition Observation Unit 422 Execution Figure 9Step S300 of the open-circuit voltage estimation method shown is as follows: The state observation unit 422 acquires the state variables of the secondary battery 404, including the terminal current Ite(t) and terminal voltage Vte(t) of the operating secondary battery 404, from the characteristic measuring unit 406 at predetermined time intervals. Thus, the state observation unit 422 obtains time-series data of the state variables measured at predetermined time intervals.

[0137] Preprocessing Unit 424 Execution Figure 9 Step S302 of the open-circuit voltage estimation method shown. That is, the preprocessing unit 424 preprocesses the state variables obtained by the state observation unit 422 and calculates the voltage estimation input data. Specifically, the preprocessing unit 424 calculates the difference between the terminal current Ite (i.e., current difference δIte) and the difference between the terminal voltage Vte (i.e., voltage difference δVte) based on the time series data of the terminal current Ite and the terminal voltage Vte obtained by the state observation unit 422. Figure 11 The processing step 500 is shown. Then, the preprocessing unit 424 calculates the rate of change of the voltage difference δVte relative to the current difference δIte, i.e., the differential gradient Sdiff. Figure 11 The process shown is 502).

[0138] Then, the preprocessing unit 424 generates voltage estimation input data, which includes terminal current Ite, terminal voltage Vte, and the time series data of each of the calculated Sdiff, from the past tracing back to a predetermined time T2 up to the present. Figure 11 (Processing 504).

[0139] Next, the state estimation unit 426 performs... Figure 9 Step S304 of the open-circuit voltage estimation method shown. That is, the state estimation unit 426 uses the learned open-circuit voltage estimation model to estimate the open-circuit voltage Voc of the secondary battery 404 in operation based on the voltage estimation input data generated above. Figure 11 (Processing 506).

[0140] Then, the state estimation unit 426 uses, for example, time series data of the open-circuit voltage Voc, terminal voltage Vte, and terminal current Ite from the past, traced back to a predetermined time T3, up to the present as input, to estimate the SOC and SOH as the current state of the secondary battery 404 according to the prior art. Figure 11 (Processing 508). For example, the state estimation unit 426 can calculate the current SOC of the secondary battery 404 based on the SOC-OCV characteristics of the secondary battery 404 and the Voc contained in the state estimation input data mentioned above.

[0141] Furthermore, the present invention is not limited to the structure of the above-described embodiments, and can be implemented in various ways without departing from its spirit.

[0142] For example, in the first and second embodiments, the current difference δIte and the voltage difference δVte are respectively the fourth-order difference Δ of the terminal current. 4 The fourth-order difference Δ between Ite and the terminal voltage 4 Vte, but the current difference δIte and voltage difference δVte do not necessarily have to be fourth-order differences. Even if the current difference δIte and voltage difference δVte are, for example, first-order differences Δ 1 Ite and Δ 1 Vte also enables the open-circuit voltage estimation model 124 to learn the relationship between the change (gradient) of terminal current relative to terminal voltage and the open-circuit voltage. However, higher-order differences (fourth order or above) can extract more common patterns of terminal current and terminal voltage changes among secondary batteries with different electrical characteristics. Therefore, it is preferable from the viewpoint of estimating the open-circuit voltage of secondary batteries from different manufacturers or models with higher accuracy.

[0143] Furthermore, time-series data of the temperature of the secondary battery 102 can be appended to the voltage estimation input data. This further improves the estimation accuracy of the open-circuit voltage based on the open-circuit voltage estimation model 124.

[0144] Furthermore, in the above embodiment, the open-circuit voltage estimation model 124 is an RNN that can easily process continuous data in the time series as input, but the structure of the open-circuit voltage estimation model is not limited to RNN.

[0145] For example, the open-circuit voltage estimation model 124 can also be constructed using a one-dimensional CNN (Convolutional Neural Network). In this case, voltage estimation input data represented by a second-order tensor (Equation (1)) can also be input into the open-circuit voltage estimation model 124.

[0146] Furthermore, in the second embodiment described above, as an example of an apparatus for performing step S304 for estimating the open-circuit voltage of a secondary battery in operation, a state estimation apparatus 400 for estimating the state of a secondary battery 404 mounted on a vehicle 402 in operation is shown. However, step S304 for estimating the open-circuit voltage of a secondary battery in operation is not limited to secondary batteries used in vehicles, and can be used for state estimation of secondary batteries used in any application such as mobile phones, bicycles, and homes.

[0147] Furthermore, in the second embodiment described above, Figure 9The open-circuit voltage estimation method shown is performed in the state estimation device 400 that performs state estimation of the secondary battery. However, this is an example; the open-circuit voltage estimation method can also be performed in a separate device that only performs open-circuit voltage estimation. Alternatively, the open-circuit voltage estimation method can also be performed in a device with various other functions. For example, Figure 9 The open-circuit voltage estimation method shown can be executed in a controller that controls the load of the secondary battery. As a specific example, for instance, in... Figure 10 In this case, the state observation unit 422, preprocessing unit 424, and state estimation unit 426 of the processing unit 420 of the state estimation device 400 can also be implemented in the processing unit 440 of the driving control device 414. In this case, the open-circuit voltage estimation model 430 stored in the storage device 428 is stored in the storage device 448 of the driving control device 414.

[0148] As explained above, the learning method for the open-circuit voltage estimation model in the first embodiment includes the following step S100: measuring the state variables, including terminal current Ite and terminal voltage Vte, of the secondary battery 102 connected to the load 106 or charger 104 at predetermined time intervals. Furthermore, the learning method includes the following steps: preprocessing the aforementioned state variables and calculating voltage estimation input data. V x1(t n Step S102; and the open-circuit voltage estimation model 124 learns the voltage estimation input data through machine learning. V x1(t n Step S104 involves the relationship between the voltage input data and the open-circuit voltage Voc of the secondary battery 102. Furthermore, this step calculates the voltage estimation input data. V x1(t n In step S102, based on the time series data of the terminal current Ite and the terminal voltage Vte, the difference between the terminal current Ite (i.e., the current difference δIte) and the difference between the terminal voltage Vte (i.e., the voltage difference δVte) are calculated (S200). Furthermore, in step S102, based on the time series data of the current difference δIte and the voltage difference δVte, the rate of change of the voltage difference δVte relative to the current difference δIte, i.e., the differential gradient Sdiff, is calculated from the past (after tracing back to the first predetermined time T1) to the present (S202). Moreover, in step S102, voltage estimation input data including the time series data of the terminal current Ite, the time series data of the terminal voltage Vte, and the time series data of the differential gradient Sdiff is generated. V x1(t n (S204).

[0149] Based on this learning method, the open-circuit voltage of secondary batteries with various electrical characteristics from different manufacturers or models can be estimated with high accuracy through the generated open-circuit voltage estimation model.

[0150] Furthermore, the differential gradient Sdiff is calculated using the least squares method. This structure prevents a decrease in the accuracy of the open-circuit voltage estimation due to measurement errors of state variables in the generated open-circuit voltage estimation model.

[0151] Furthermore, the current difference δIte and voltage difference δVte are the fourth-order difference Δ of the time series data of the terminal current Ite. 4 Fourth-order difference Δ of time series data of Ite and terminal voltage Vte 4 Vte. Based on this structure, the higher-order variation patterns of terminal current Ite and terminal voltage Vte, which are more commonly found in secondary batteries with different electrical characteristics, can be more accurately estimated using an open-circuit voltage estimation model.

[0152] Furthermore, the open-circuit voltage estimation model 124 is constructed from an RNN or a one-dimensional CNN. Additionally, the intermediate layers of the RNN constituting the open-circuit voltage estimation model 124 can be constructed from LSTM or GRU. Based on this structure, time-series data with multiple variables can be processed effectively, enabling efficient learning of the open-circuit voltage estimation model.

[0153] Furthermore, the open-circuit voltage estimation model 124 is generated through learning using time-series data of state variables, including terminal current Ite and terminal voltage Vte, of multiple secondary batteries 102 with different electrical characteristics connected to a load 106 or a charger 104. Based on this structure, an open-circuit voltage estimation model capable of accurately estimating the open-circuit voltage of secondary batteries with various electrical characteristics from different manufacturers or models can be generated.

[0154] Furthermore, the open-circuit voltage estimation method for the secondary battery in the second embodiment described above includes the following steps: step S300, measuring the state variables of the secondary battery in operation, including terminal current and terminal voltage, at predetermined time intervals; and step S302, preprocessing the state variables and calculating voltage estimation input data. Additionally, this open-circuit voltage estimation method includes the following step S304: using the open-circuit voltage estimation model learned through the learning method shown in the first embodiment, estimating the open-circuit voltage of the secondary battery in operation based on the voltage estimation input data. Moreover, in step S302, which calculates the voltage estimation input data, the difference between the terminal current (i.e., current difference) and the difference between the terminal voltage (i.e., voltage difference) are calculated based on the time series data of the terminal current and the time series data of the terminal voltage. Figure 11 The processing step 500). Furthermore, in step S302, based on the time-series data of the current difference and the voltage difference, the rate of change of the voltage difference relative to the current difference, i.e., the differential gradient Sdiff, is calculated over the period from the time before the first predetermined time T1 was traced back to the present. Figure 11 Processing 502). Furthermore, in step S302, voltage estimation input data is generated, which includes time-series data of terminal current, time-series data of terminal voltage, and time-series data of differential gradient (…). Figure 11 (Processing 504).

[0155] Based on this structure, it is possible to estimate the open-circuit voltage of secondary batteries with various electrical characteristics from different manufacturers or models with high accuracy during operation.

[0156] Furthermore, the state estimation apparatus 400 of the second embodiment includes: a state observation unit 422, which measures state variables including terminal current and terminal voltage of the secondary battery 404 in operation at predetermined time intervals; a preprocessing unit 424, which preprocesses the state variables measured by the state observation unit 422 and calculates input data; and a state estimation unit 426, which estimates the current charging rate and / or degradation degree of the secondary battery 404 based on the calculated input data. Moreover, the state estimation unit 426 uses the open-circuit voltage estimation model 430 learned by the learning method of the first embodiment to estimate the current open-circuit voltage of the secondary battery 404, and uses the estimated open-circuit voltage to estimate the current charging rate and / or degradation degree of the secondary battery 404. According to this structure, the open-circuit voltage of secondary batteries with various electrical characteristics from different manufacturers or models can be estimated with high accuracy during operation, thereby estimating the state of these secondary batteries with high accuracy.

[0157] [Structure supported by the above embodiments]

[0158] The above implementation supports the following structures.

[0159] (Structure 1) A learning method for an open-circuit voltage estimation model of a secondary battery, which is a machine learning-based learning method for estimating the open-circuit voltage of a working secondary battery connected to a load or charger. The learning method for the open-circuit voltage estimation model of the secondary battery includes the following steps: measuring state variables including terminal current and terminal voltage of the working secondary battery at predetermined time intervals; preprocessing the state variables and calculating voltage estimation input data; and using machine learning to enable the open-circuit voltage estimation model to learn the voltage estimation input data and the open-circuit voltage of the working secondary battery. In the calculation step, based on the time series data of the terminal current and the time series data of the terminal voltage, the difference between the terminal current (i.e., current difference) and the difference between the terminal voltage (i.e., voltage difference) are calculated. Based on the time series data of the current difference and the time series data of the voltage difference, the rate of change of the voltage difference relative to the current difference (i.e., differential gradient) is calculated from the past, which traces back to the first predetermined time, to the present. This generates the voltage estimation input data, which includes the time series data of the terminal current, the time series data of the terminal voltage, and the time series data of the differential gradient.

[0160] (Structure 2) In the learning method of the open-circuit voltage estimation model of the secondary battery described in Structure 1, the differential gradient is calculated using the least squares method.

[0161] (Structure 3) In the learning method of the open-circuit voltage estimation model of the secondary battery described in Structure 1 or 2, the current difference and the voltage difference are respectively the fourth-order difference of the time series data of the terminal current and the fourth-order difference of the time series data of the terminal voltage.

[0162] (Structure 4) In the learning method of the open-circuit voltage estimation model of the secondary battery described in any of Structures 1 to 3, the open-circuit voltage estimation model is composed of a recurrent neural network (RNN).

[0163] (Structure 5) In the learning method of the open-circuit voltage estimation model of the secondary battery described in Structure 4, the intermediate layer of the RNN constituting the open-circuit voltage estimation model is composed of Long Short-Term Memory (LSTM) or Gated Recursive Unit (GRU).

[0164] (Structure 6) In the learning method of the open-circuit voltage estimation model of the secondary battery described in any of Structures 1 to 3, the open-circuit voltage estimation model is composed of a one-dimensional convolutional neural network (CNN).

[0165] (Structure 7) In the learning method of the open-circuit voltage estimation model of the secondary battery described in any of Structures 1 to 6, the open-circuit voltage estimation model is generated by learning using time series data of state variables including terminal current and terminal voltage of each of the multiple secondary batteries with different electrical characteristics connected to a load or charger.

[0166] (Structure 8) A method for estimating the open-circuit voltage of a secondary battery, comprising the following steps: measuring state variables of the secondary battery in operation, including terminal current and terminal voltage, at predetermined time intervals; preprocessing the state variables and calculating voltage estimation input data; and estimating the open-circuit voltage of the secondary battery in operation based on the voltage estimation input data using the open-circuit voltage estimation model learned by any one of Structures 1 to 7, which has been trained by the learning method of the open-circuit voltage estimation model of the secondary battery described in any one of Structures 1 to 7. In the calculation step, the difference between the terminal current (i.e., current difference) and the difference between the terminal voltage (i.e., voltage difference) are calculated based on the time series data of the terminal current and the time series data of the terminal voltage. Based on the time series data of the current difference and the time series data of the voltage difference, the rate of change of the voltage difference relative to the current difference (i.e., differential gradient) is calculated from the past, which traces back to the present, to the present. The voltage estimation input data, comprising the time series data of the terminal current, the time series data of the terminal voltage, and the time series data of the differential gradient, is generated.

[0167] (Structure 9) A state estimation device for a secondary battery, which estimates the state of a secondary battery in operation, wherein the state estimation device comprises: a state observation unit that measures state variables including terminal current and terminal voltage of the secondary battery in operation at predetermined time intervals; a preprocessing unit that preprocesses the state variables measured by the state observation unit and calculates input data; and a state estimation unit that estimates the current charge rate and / or degradation of the secondary battery in operation based on the input data, wherein the state estimation unit estimates the current open-circuit voltage of the secondary battery in operation using an open-circuit voltage estimation model learned by a learning method for an open-circuit voltage estimation model of a secondary battery described in any one of Structures 1 to 7, and uses the estimated open-circuit voltage to estimate the current charge rate and / or degradation of the secondary battery in operation.

Claims

1. A learning method for an open-circuit voltage estimation model of a secondary battery, which is a machine learning-based learning method for estimating the open-circuit voltage of a working secondary battery connected to a load or charger, wherein... The learning method for the open-circuit voltage estimation model of the secondary battery includes the following steps: The state variables, including terminal current and terminal voltage, of the secondary battery in operation are measured at specified time intervals. The state variables are preprocessed and voltage estimation input data is calculated; as well as Through machine learning, the open-circuit voltage estimation model learns the relationship between the voltage estimation input data and the open-circuit voltage of the operating secondary battery. In the step of performing the aforementioned calculation Based on the time series data of the terminal current and the time series data of the terminal voltage, calculate the difference between the terminal current (current difference) and the difference between the terminal voltage (voltage difference). Based on the time series data of the current difference and the time series data of the voltage difference, the rate of change of the voltage difference relative to the current difference, i.e., the differential gradient, is calculated from the time elapsed since the first predetermined time to the present. Generate voltage estimation input data that includes time-series data of the terminal current, time-series data of the terminal voltage, and time-series data of the differential gradient. The current difference and the voltage difference are the fourth-order differences of the time series data of the terminal current and the fourth-order differences of the time series data of the terminal voltage, respectively.

2. The learning method for the open-circuit voltage estimation model of a secondary battery according to claim 1, wherein, The differential gradient is calculated using the least squares method.

3. The learning method for the open-circuit voltage estimation model of a secondary battery according to claim 1, wherein, The open-circuit voltage estimation model is composed of a recurrent neural network (RNN).

4. The learning method for the open-circuit voltage estimation model of a secondary battery according to claim 3, wherein, The intermediate layers of the RNN constituting the open-circuit voltage estimation model are composed of Long Short-Term Memory (LSTM) or Gated Recursive Units (GRU).

5. The learning method for the open-circuit voltage estimation model of a secondary battery according to claim 1, wherein, The open-circuit voltage estimation model is composed of a one-dimensional convolutional neural network (CNN).

6. The learning method for the open-circuit voltage estimation model of a secondary battery according to any one of claims 1 to 5, wherein, The open-circuit voltage estimation model is generated by learning from time-series data of state variables, including terminal current and terminal voltage, of multiple secondary batteries with different electrical characteristics connected to a load or charger.

7. A method for estimating the open-circuit voltage of a secondary battery, wherein, The method for estimating the open-circuit voltage of the secondary battery includes the following steps: The state variables, including terminal current and terminal voltage, of the secondary battery in operation are measured at specified time intervals. The state variables are preprocessed and voltage estimation input data is calculated; as well as The open-circuit voltage estimation model, learned using the learning method for the open-circuit voltage estimation model of a secondary battery according to any one of claims 1 to 6, is used to estimate the open-circuit voltage of the secondary battery in operation based on the voltage estimation input data. In the step of performing the aforementioned calculation Based on the time series data of the terminal current and the time series data of the terminal voltage, calculate the difference between the terminal current (current difference) and the difference between the terminal voltage (voltage difference). Based on the time series data of the current difference and the time series data of the voltage difference, the rate of change of the voltage difference relative to the current difference, i.e., the differential gradient, is calculated from the time elapsed since the first predetermined time to the present. Generate voltage estimation input data that includes time-series data of the terminal current, time-series data of the terminal voltage, and time-series data of the differential gradient.

8. A state estimation device for a secondary battery, which is a state estimation device for estimating the state of a secondary battery during operation, wherein, The state estimation device for the secondary battery includes: The status observation unit measures the status variables, including terminal current and terminal voltage, of the secondary battery in operation at predetermined time intervals. The preprocessing unit preprocesses the state variables measured by the state observation unit and calculates the input data; as well as The state estimation unit estimates the current charge rate and / or degradation of the operating secondary battery based on the input data. The state estimation unit uses an open-circuit voltage estimation model learned by the learning method of the open-circuit voltage estimation model of the secondary battery according to any one of claims 1 to 6 to estimate the current open-circuit voltage of the secondary battery in operation, and uses the estimated open-circuit voltage to estimate the current charging rate and / or degradation of the secondary battery in operation.

Images