Learning method for state estimation model, manufacturing method for state estimation device, state estimation system and state estimation device

Abstract

To provide a state estimation model that enables the rapid estimation of battery status compared to learning a parameter for a physical model.SOLUTION: PU generates a battery model for a battery pack by combining n battery cell models specified by cell model data. PU sets model specification data that defines the battery model to values representing a normal battery pack and values representing an abnormal battery pack (S10, S12). PU generates training data for an identification model using simulations based on the battery model. The identification model has the battery pack current and voltage as input variables and outputs whether the battery pack is degraded. PU trains the identification model using the training data (S20).SELECTED DRAWING: Figure 3

Description

Technical Field

[0001] The present invention relates to a method for learning a state estimation model, a method for manufacturing a state estimation device, a state estimation system, and a state estimation device.

Background Art

[0002] For example, Patent Document 1 below describes an apparatus for estimating spots of the salt concentration of a secondary battery. Specifically, this apparatus uses a physical model of the secondary battery for estimating the salt concentration. In particular, this apparatus includes a process of calculating parameters defining the physical model by the least squares method based on measurement values such as the voltage of the secondary battery as input variables.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] As described above, in order to calculate the parameters defining the physical model based on the detected values of the state of the actual secondary battery, it is necessary to use the detected values over a relatively long period of time. Therefore, it is difficult to satisfy the requirement of quickly estimating the state of the battery based on the input variables.

Means for Solving the Problems

[0005] Hereinafter, means for solving the above problems and their effects will be described. 1. A method for training a state estimation model for estimating the state of a battery, the method comprising steps of performing a model definition process, a training data generation process, and a training process, wherein the model definition process is a process of setting defined parameters, which are parameters defining an input-output model, to values representing respective states of different batteries, the input-output model is a model in which either one of the current and voltage of the battery is included in the input and the other is included in the output, the training data generation process is a process of generating training data for the state estimation model, the training data includes data of input and output obtained by simulation using the input-output model and data indicating the state of the battery represented by the values of the defined parameters of the input-output model, the training process is a process of training the state estimation model with the training data, the input variables of the state estimation model include the input and output of the input-output model, and the output of the state estimation model is a method for training a state estimation model that indicates the state of the battery.

[0006] According to the state estimation model as the learned model described above, the state of the battery can be estimated more quickly compared to the case of learning the parameters of a physical model. Furthermore, in the above method, the state of the battery is represented by the defined parameters of the input-output model. Then, by simulation using each input-output model in which the state of the battery is represented, the behavior of the current and voltage of the battery according to the state of the battery can be obtained. Therefore, in the above method, the behavior of the current and voltage of the battery according to the state of the battery can be obtained without charging and discharging using an actual battery having a desired state. And by using them as training data, a state estimation model can be generated. Therefore, according to the above method, training data for generating a state estimation model can be easily obtained.

[0007] 2. The battery is an assembled battery in which a plurality of battery cells are connected in series, the input / output model is a model in which a plurality of cell models are connected in series, the cell model is a model in which either one of the battery cells is included in the input and the other is included in the output, and the model defining process includes a process of expressing the state of the assembled battery by determining the value of the defined parameter for each cell model. The learning method of the state estimation model according to the above 1.

[0008] The above method includes a process of setting the value of a defined parameter for each cell model of the battery cells constituting the assembled battery. Therefore, the degree of freedom in expressing the state of the assembled battery can be increased as compared with the case where the values of the defined parameters of all cell models are the same.

[0009] 3. The learning method of the state estimation model according to the above 1 or 2, wherein the state estimation model is a discrimination model for estimating the presence or absence of progress of deterioration of the battery. 4. The training data used for learning the state estimation model by the above learning process includes, in addition to the training data generated by the training data generation process, data obtained by charging and discharging an actual battery and data indicating the state of the battery. The learning method of the state estimation model according to any one of the above 1 to 3.

[0010] In the above method, by including the training data obtained from the actual battery, learning closer to the state of the actual battery can be executed as compared with the case of using only the data obtained by simulation.

[0011] 5. A method for learning a state estimation model, comprising steps of executing an individual data calculation process, a training data regeneration process, and a relearning process. The individual data calculation process is a process of calculating values of the specified parameters by measuring the voltage and current of one battery. The training data regeneration process includes a process of generating data of the input and output of the input-output model and data indicating the state of the battery represented by the values of the specified parameters of the input-output model by simulation using the input-output model defined by the values of the specified parameters calculated by the individual data calculation process. The relearning process is a process of relearning the state estimation model using the training data generated by the training data regeneration process. The method for learning a state estimation model according to any one of 1 to 4 above.

[0012] The accuracy of the state discrimination model learned by the learning process depends on the training data generated by the training data generation process. Therefore, for example, even if the value of a certain specified parameter is set to a value representing a state indicating a predetermined state, if the values of other specified parameters are slightly different, the estimation results may be different. On the other hand, in the above method, since the values of other specified parameters are set using the one battery, the possibility of different estimation results can be reduced. That is, in the above method, a state estimation model capable of accurately estimating the state of the one battery can be obtained by relearning.

[0013] 6. A method for manufacturing a state estimation device for estimating the state of a battery, the state estimation device being configured to execute an input variable acquisition process and a state estimation process. The input variable acquisition process is a process of acquiring the values of the input variables of the state estimation model in the method for learning a state estimation model according to any one of 1 to 5 above. The state estimation process is a process of estimating the state of the battery by inputting the values of the input variables acquired by the input variable acquisition process into the state estimation model. A method for manufacturing a state estimation device having each step in the method for learning a state estimation model according to any one of 1 to 5 above.

[0014] In the above manufacturing method, since training data for generating a state estimation model can be easily obtained, the man-hours required for manufacturing the state estimation device can be reduced. 7. A state estimation system for estimating the state of a battery, configured to execute model specification processing, training data generation processing, learning processing, input variable acquisition processing, and state estimation processing, wherein the model specification processing is a process of setting specified parameters, which are parameters for specifying an input / output model, to values representing different states of the battery, the input / output model is a model in which either one of the current and voltage of the battery is included in the input and the other is included in the output, the training data generation processing is a process of generating training data for a state estimation model, the training data includes data of inputs and outputs obtained by simulation using the input / output model and data indicating the state of the battery represented by the values of the specified parameters of the input / output model, the learning processing is a process of learning the state estimation model with the training data, the input variables of the state estimation model include the inputs and outputs of the input / output model, the output of the state estimation model indicates the state of the battery, the input variable acquisition processing is a process of acquiring the values of the input variables of the state estimation model, and the state estimation processing is a process of estimating the state of the battery by inputting the values acquired by the input variable acquisition processing into the state estimation model.

[0015] In the above configuration, the state of the battery is represented by the specified parameters of the input / output model. Then, by simulation using each input / output model in which the state of the battery is represented, the behavior of the current and voltage of the battery according to the state of the battery can be obtained. Therefore, in the above configuration, without preparing actual batteries having a desired state and performing charge and discharge using them, the behavior of the current and voltage of the battery according to the state of the battery can be obtained. And by using them as training data, a state estimation model can be generated. Therefore, according to the above configuration, training data for generating a state estimation model can be easily obtained.

[0016] Furthermore, in the above configuration, by estimating the state of the battery using the state estimation model, the state can be estimated more quickly compared to the case where the state is estimated by obtaining parameters that define the model of the battery to be the target of the state estimation.

[0017] 8. The state estimation system according to item 7 above, comprising a state estimation device and a learning device, wherein the state estimation device is configured to execute the input variable acquisition process and the state estimation process, and the learning device is configured to execute the model definition process, the training data generation process, and the learning process.

[0018] According to the above configuration, by providing a state estimation device that executes an estimation process using a state estimation model separately from the learning device, the location where the state estimation process is executed can be freely selected. 9. The state estimation device is configured to execute an individual data calculation process, a specified parameter transmission process, and a state estimation model reception process. The learning device is configured to execute a specified parameter reception process, a training data regeneration process, a relearning process, and a state estimation model transmission process. The individual data calculation process is a process of calculating the value of the specified parameter by measuring the voltage and current of one actual battery. The specified parameter transmission process is a process of transmitting the specified parameter calculated by the individual data calculation process to the learning device. The specified parameter reception process is a process of receiving the specified parameter transmitted by the specified parameter transmission process. The training data regeneration process is a process of generating, by simulation using the input-output model defined by the value of the specified parameter received by the specified parameter reception process, data of the input and output of the input-output model and data indicating the state of the battery represented by the value of the specified parameter of the input-output model. The relearning process is a process of relearning the state estimation model using the training data generated by the training data regeneration process. The state estimation model transmission process is a process of transmitting the state estimation model relearned by the relearning process to the state estimation device. The state estimation model reception process is a process of receiving the state estimation model transmitted by the state estimation model transmission process. The state estimation system according to item 8 above.

[0019] The accuracy of the state estimation model learned by the learning process depends on the training data generated by the training data generation process. Therefore, for example, even if the value of a certain specified parameter is set to a value representing a state indicating a predetermined state, if the value of another specified parameter is slightly different, the estimation result may be different. On the other hand, in the above configuration, the state estimation model is relearned using the specified parameter of one actual battery. Thereby, a state estimation model specialized for one actual battery to be estimated can be obtained.

[0020] 10. The state estimation device in the state estimation system according to item 9 above.

Brief Description of the Drawings

[0021]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Embodiments for Carrying Out the Invention

[0022] Hereinafter, an embodiment will be described with reference to the drawings. 「Premise Configuration」 FIG. 1 shows the configuration of a state estimation system for a battery pack.

[0023] The vehicle 2 includes a motor generator 4. The rotating shaft of the motor generator 4 is mechanically connected to the drive wheels of the vehicle. The output voltage of the power conversion circuit 6 is applied to the terminals of the motor generator 4. The power conversion circuit 6 is configured to convert the terminal voltage of the battery pack 10 as a DC voltage source into an AC voltage and output it.

[0024] The battery pack 10 is a series connection of battery cells 12(1), 12(2),... 12(n). The terminal voltage of the battery pack 10 may be, for example, several tens of volts to several hundreds of volts. The numbers in parentheses in the battery cells 12(1), 12(2),... 12(n) are numbers for identifying the individual units. Hereinafter, when collectively referring to the battery cells 12(1), 12(2),... 12(n), they will be described as battery cells 12. The battery cell 12 is, as an example, a lithium-ion secondary battery. The battery cell 12 has, as an example, a structure in which a positive electrode plate, a separator, and a negative electrode plate are laminated and flatly wound.

[0025] The monitoring unit 20 is a circuit that monitors the states of the battery cells 12(1), 12(2), …, 12(n) of the assembled battery 10. The battery ECU 30 includes a PU 32, a storage device 34, and a communication device 36. The PU 32 is a software processing device such as a CPU and a GPU. The storage device 34 may be an electrically rewritable non-volatile memory and a storage medium such as a disk medium. The battery ECU 30 is a device that monitors the state of the assembled battery 10.

[0026] The battery ECU 30 refers to the charge and discharge current I of the assembled battery 10 detected by the current sensor 22. Also, the battery ECU 30 refers to the cell voltages Vc(1), Vc(2), …, Vc(n), which are the respective terminal voltages of the battery cells 12(1), 12(2), …, 12(n), detected by the monitoring unit 20. Further, the battery ECU 30 refers to the temperature T of the assembled battery 10 detected by the monitoring unit 20.

[0027] The battery ECU 30 can communicate with the upper ECU 40 via the in-vehicle network. The upper ECU 40 is a device that manages the power of the drive system of the vehicle. The upper ECU 40 controls the driving force of the motor generator 4 by outputting a command to the MG ECU 44 via the in-vehicle network. By controlling the driving force, the charge and discharge power amount of the assembled battery 10 is also controlled. Therefore, when the upper ECU 40 outputs a command to the MG ECU 44 via the in-vehicle network, the charge and discharge power amount of the assembled battery 10 is controlled.

[0028] The MG ECU 44 operates the power conversion circuit 6 to control the torque or the like, which is the control amount of the motor generator 4 as the control target. The battery ECU 30 sets the maximum value Imax of the charging current at which lithium does not precipitate at the negative electrode of the assembled battery 10 and outputs it to the upper ECU 40. The upper ECU 40 calculates the maximum charging power Winmax, which is the maximum value of the charging power of the assembled battery 10, based on the maximum value Imax and outputs it to the MG ECU 44. The battery ECU 30 sets the maximum value Imax according to the presence or absence of deterioration of the assembled battery 10. The battery ECU 30 determines the presence or absence of deterioration of the assembled battery 10 based on an identification model defined by the identification model data 34a stored in the storage device 34.

[0029] The battery ECU 30 can communicate with a learning device 60 outside the vehicle 2 via a wireless network 50. The learning device 60 can communicate with a plurality of vehicles 2. The learning device 60 includes a PU 62, a storage device 64, and a communication device 66. The PU 62 is a software processing device such as a CPU and a GPU. The storage device 64 may be an electrically rewritable non-volatile memory and a storage medium such as a disk medium. The storage device 64 stores cell model data 64a that defines a cell model, which is a model of the battery cell 12. The learning device 60 generates training data for generating the identification model data 34a using the cell model.

[0030] This will be described in detail below. "Cell Model" Fig. 2 shows the cell model.

[0031] The negative electrode solid-phase diffusion model M10 is a process of calculating the lithium ion concentration on the surface of the negative electrode of the battery cell 12 using a mathematical model based on the charge and discharge current I and the temperature T as input variables.

[0032] The negative electrode potential map M12 is a process of calculating the negative electrode potential of the battery cell 12 based on the lithium ion concentration on the surface of the negative electrode as an input variable. Specifically, the negative electrode potential map M12 includes a process of calculating the negative electrode active material SOC by the following formula based on the lithium ion concentration on the surface of the negative electrode as an input variable.

[0033] Negative electrode active material SOC = {Lithium ion concentration on the surface of the negative electrode} / cnmax In the above formula, the maximum lithium ion concentration cnmax of the negative electrode as a specified parameter for defining the model is used.

[0034] The negative electrode potential map M12 includes a process of calculating the negative electrode potential based on the SOC of the negative electrode active material as an input variable. Specifically, this process is, for example, a process of performing a map operation on the negative electrode potential by the PU62 with the map data stored in the storage device 64. The map data is data in which the SOC of the negative electrode active material is an input variable and the negative electrode potential is an output variable.

[0035] Here, the map data is a set of data of discrete values of the input variable and values of the output variable corresponding to each value of the input variable. Also, the map operation may be a process in which when the value of the input variable matches any of the values of the input variable of the map data, the value of the output variable of the corresponding map data is the operation result. Also, the map operation may be a process in which when the value of the input variable does not match any of the values of the input variable of the map data, the value obtained by interpolating the values of a plurality of output variables included in the map data is the operation result. Alternatively, the map operation may be a process in which when the value of the input variable does not match any of the values of the input variable of the map data, the value of the output variable of the map data corresponding to the closest value among the values of a plurality of input variables included in the map data is the operation result.

[0036] The negative electrode reaction overvoltage model M14 is a process of calculating the negative electrode reaction overvoltage, which is an overvoltage that can occur at the negative electrode of the battery cell 12, using a mathematical model based on the charge and discharge current I and the temperature T as input variables.

[0037] The negative electrode potential calculation process M16 is a process of substituting, into the estimated value of the negative electrode potential, a value obtained by adding the negative electrode reaction overvoltage to the negative electrode potential calculated by the negative electrode potential map M12. The positive electrode solid-phase diffusion model M18 is a process of calculating the lithium-ion concentration on the surface of the positive electrode of the battery cell 12 using a mathematical model based on the charge and discharge current I and the temperature T as input variables.

[0038] The positive electrode potential map M20 includes a process of calculating the positive electrode potential using a mathematical model based on the lithium-ion concentration on the surface of the positive electrode as an input variable. Specifically, the positive electrode potential map M20 includes a process of calculating the state of charge (SOC) of the positive electrode active material according to the following formula based on the lithium-ion concentration on the surface of the positive electrode as an input variable.

[0039] State of charge of positive electrode active material = 1 - {lithium-ion concentration on the surface of the positive electrode} / cpmax In the above formula, the maximum lithium-ion concentration cpmax of the positive electrode as a defining parameter for defining the model is used.

[0040] The positive electrode potential map M20 includes a process of calculating the positive electrode potential based on the state of charge (SOC) of the positive electrode active material as an input variable. Specifically, as an example, this process is a process of performing a map operation on the positive electrode potential by the PU62 with the map data stored in the storage device 64. The map data is data with the charge and discharge current I and the temperature T as input variables and the positive electrode potential as an output variable.

[0041] The positive electrode reaction overvoltage model M22 is a process of calculating the positive electrode reaction overvoltage, which is the overvoltage that can occur at the positive electrode of the battery cell 12, using a mathematical model based on the charge and discharge current I and the temperature T as input variables.

[0042] The positive electrode potential calculation process M24 is a process of substituting the value obtained by adding the positive electrode reaction overvoltage to the positive electrode potential calculated by the positive electrode potential map M20 into the estimated value of the positive electrode potential. The liquid-phase diffusion model M26 is a process of calculating the salt concentration in the electrolyte in the thickness direction of the battery cell 12 using a mathematical model based on the charge and discharge current I and the temperature T as input variables. Here, the mathematical model is, as an example, a liquid-phase diffusion model. Specifically, the liquid-phase diffusion model M26 includes a process of performing a mapping operation on the negative electrode boundary salt concentration and the positive electrode boundary salt concentration based on the state of charge SOC and the temperature T as input variables. The negative electrode boundary salt concentration is the salt concentration at the boundary between the negative electrode and the negative electrode current collector. Also, the positive electrode boundary salt concentration is the salt concentration at the boundary between the positive electrode and the positive electrode current collector. The liquid-phase diffusion model M26 includes a process of calculating the salt concentration in the electrolyte in the thickness direction of the battery cell 12 using a liquid-phase diffusion model based on the charge and discharge current I and the temperature T as input variables, with the negative electrode boundary salt concentration and the positive electrode boundary salt concentration as boundary conditions. The thickness direction of the battery cell 12 is the direction parallel to the shortest side when the electrode body in which the positive electrode plate, the separator, and the negative electrode plate are laminated and wound is regarded as a rectangular parallelepiped.

[0043] The liquid-phase potential difference formula M28 is a process of calculating the liquid-phase potential difference, which is the potential difference in the electrolyte in the thickness direction of the battery cell 12, based on the salt concentration in the above-mentioned electrolyte as an input variable. The component resistance model M30 is a process of calculating the voltage drop amount due to the resistance of the components included in the battery cell 12 using a mathematical model based on the charge and discharge current I and the temperature T as input variables. The component resistance model M30 includes a process of calculating the component resistance R using the proportional coefficient α and the intercept b as, for example, "R = α·T + b". The component resistance model M30 includes a process of substituting the value obtained by multiplying the component resistance R by the charge and discharge current I into the voltage drop amount due to the resistance of the components.

[0044] The liquid resistance model M32 is a process of calculating the voltage drop amount in the electrolyte using a mathematical model based on the voltage drop amount due to the resistance of the components as an input variable. The film resistance model M34 is a process of calculating the voltage drop amount due to the film formed on the negative electrode surface using a mathematical model based on the voltage drop amount in the electrolyte as an input variable.

[0045] The voltage calculation process M40 is a process of substituting, into the voltage estimated value, a value obtained by subtracting the sum of the estimated value of the negative electrode potential and the voltage drop amount due to the film from the sum of the estimated value of the positive electrode potential and the liquid phase potential difference. In the cell model data 64a, the specified parameters, which are the parameters defining the cell model, are set to match the behavior of the central characteristic product of the battery cell 12. That is, for example, the maximum negative electrode lithium ion concentration cnmax, the maximum positive electrode lithium ion concentration cpmax, the proportional coefficient α, the intercept b, etc. are set according to the central characteristic product.

[0046] "Generation process of the identification model" Fig. 3 shows the procedure of the generation process of the identification model. The process shown in Fig. 3 is realized by the PU62 repeatedly executing, for example, at a predetermined cycle, the program stored in the storage device 64. Hereinafter, the step numbers of each process are represented by numbers preceded by "S".

[0047] In the series of processes shown in Fig. 3, the PU62 first sets the specified parameters defining the battery model of the normal assembled battery 10 (S10). The battery model of the assembled battery 10 is a model in which n cell models of the battery cell 12 shown in Fig. 2 are connected in series. The battery model of the normal assembled battery 10 may be defined only by the parameters defined by the cell model data 64a, for example. That is, the specified parameters of each of the n battery cells 12 may all be values for simulating the behavior of the central characteristic product of the battery cell 12. Also, the battery model of the normal assembled battery 10 may include specified parameters with a change amount from the central characteristic product of a predetermined value or less for at least one of the battery cells 12(1) to 12(n), for example. Here, the predetermined value is set to a value regarded as having no deterioration. It is desirable that there are a plurality of types of battery models of the normal assembled battery 10. In the process of S10, the PU62 sets the variation in the state of charge SOC of each of the n battery cells 12 to a value within the allowable range.

[0048] Next, the PU62 sets the specified parameters that define the battery model of the abnormal battery pack 10 (S12). The specified parameters that define the battery model of the abnormal battery pack 10 include the specified parameters for at least one of the battery cells 12(1) to 12(n) where the amount of change from the central characteristic product is equal to or greater than the specified value. The specified value is set to a value that can be regarded as the progress of deterioration. It is desirable that there are multiple types of battery models for the abnormal battery pack 10.

[0049] In the process of S12, as an example, the PU62 represents the abnormal battery pack 10 by the maximum negative electrode lithium ion concentration cnmax, the maximum positive electrode lithium ion concentration cpmax, the proportional coefficient α, the intercept b, and the state of charge SOC of each battery cell 12.

[0050] That is, the PU62 represents the abnormal battery pack 10 by setting so that the variation of the n battery cells 12 with respect to the full charge capacity determined from the maximum negative electrode lithium ion concentration cnmax and the maximum positive electrode lithium ion concentration cpmax is out of the allowable range. Also, the PU62 represents the abnormal battery pack 10 by setting the full charge capacity determined from the maximum negative electrode lithium ion concentration cnmax and the maximum positive electrode lithium ion concentration cpmax to be equal to or less than a predetermined value. Here, the number of battery cells 12 with a full charge capacity equal to or less than the predetermined value is one or more and n or less. The PU62 may set the maximum negative electrode lithium ion concentration cnmax and the maximum positive electrode lithium ion concentration cpmax so as to cover all patterns where the number of battery cells 12 with a full charge capacity equal to or less than the predetermined value is from 1 to n.

[0051] Further, PU62 represents an abnormal battery pack 10 by setting the variation in the state of charge (SOC) of each of the n battery cells 12 to be outside the allowable range. Also, PU62 represents an abnormal battery pack 10 by a battery cell 12 in which the component resistance R determined by the proportionality coefficient α and the intercept b is equal to or greater than a threshold value. Here, the number of battery cells 12 in which the component resistance R is equal to or greater than the threshold value is one or more and n or less. PU62 may set the proportionality coefficient α and the intercept b so as to cover all patterns in which the number of battery cells 12 in which the component resistance R is equal to or greater than the threshold value is from 1 to n. Also, PU62 represents an abnormal battery pack 10 by setting the variation in the component resistance R determined by the proportionality coefficient α and the intercept b to be outside the allowable range.

[0052] PU62 simulates the terminal voltage of the battery pack 10 by inputting a current and a temperature according to a predetermined pattern to each of a normal battery model and an abnormal battery model for the battery pack 10 (S14).

[0053] PU62 stores, in the storage device 64, each set of the current, temperature, terminal voltage of the battery cell 12, and terminal voltage of the battery pack 10 obtained by the process of S14, in association with a label variable indicating whether it is data generated from a normal model (S16). Each set of the label variable and the data associated with the same label variable constitutes one piece of training data.

[0054] Note that the input variables of the identification model do not necessarily include only the value at one timing for the current of the battery cell 12, the terminal voltage of the battery cell 12, and the terminal voltage of the battery pack 10. For example, the input variables of the identification model may include time-series data consisting of values at different timings for the current of the battery cell 12, the terminal voltage of the battery cell 12, and the terminal voltage of the battery pack 10. In that case, the training data includes the label variable, the time-series data associated with the same label variable, and the temperature.

[0055] Further, the PU 62 acquires training data obtained from the actual assembled battery 10 from the outside and stores it in the storage device 64 (S18). This training data may include the current, terminal voltage, and temperature when the central characteristic product of the assembled battery 10 is actually charged and discharged, and the terminal voltage of the battery cell 12 constituting the assembled battery 10. Further, this training data may include the current, terminal voltage, and temperature when the intentionally deteriorated assembled battery 10 is actually charged and discharged, and the terminal voltage of the battery cell 12 constituting the assembled battery 10.

[0056] Next, the PU 62 learns an identification model using the training data stored in the storage device 64 by the processes of S16 and S18 (S20). Here, in practice, for example, only a part of the data stored in the storage device 64 by the processes of S16 and S18 may be used as training data, and the remaining data may be used as verification data and test data. Also, for example, by performing cross-validation, a process of changing the data used as training data among the data stored in the storage device 64 by the processes of S16 and S18 may be included.

[0057] The identification model may be a non-parametric model such as a support vector machine. In that case, the identification model data 34a defining the identification model may be support vectors extracted from the training data. Also, the identification model may be a parametric model such as a neural network. In that case, the identification model data 34a may be the weight coefficients and biases of the neural network.

[0058] Note that the PU 62 learns the identification model so that the output of the identification model matches the label variable by using a well-known cross-entropy or the like as an evaluation function. That is, for example, when the identification model is a neural network, the PU 62 updates the weight coefficients and bias coefficients by the error backpropagation method or the like based on the evaluation function.

[0059] When the learning is completed, PU62 stores the identification model data 34a in the storage device 64 (S22). The identification model data 34a thus stored in the storage device 64 is stored in the storage device 34 mounted on each vehicle.

[0060] Note that when the process of S22 is completed, PU62 temporarily ends the series of processes shown in FIG. 3. "Regarding the power limit restriction process" FIG. 4 shows the procedure of the process regarding the power limit restriction executed by the battery ECU 30. The series of processes shown in FIG. 4 is realized by the PU32 repeatedly executing the program stored in the storage device 34, for example, at a predetermined cycle.

[0061] In the series of processes shown in FIG. 4, first, PU32 acquires the charge and discharge current I, temperature T, terminal voltage V of the battery pack 10, and cell voltages Vc(1) to Vc(n) as input variables of the identification model (S30). Here, the terminal voltage V of the battery pack 10 may be a value obtained by adding the cell voltages Vc(1) to Vc(n) by the PU32. Note that the charge and discharge current I, temperature T, terminal voltage V of the battery pack 10, and cell voltages Vc(1) to Vc(n) acquired by the process of S30 are not limited to the sampling values at one timing. At least one of the variables of the charge and discharge current I, temperature T, terminal voltage V of the battery pack 10, and cell voltages Vc(1) to Vc(n) acquired by the process of S30 may be sampling values at a plurality of timings. That is, for example, when the input variables of the identification model include the time series data of the charge and discharge current I, terminal voltage V, and cell voltages Vc(1) to Vc(n), the process of S30 is a process of acquiring the synchronous series data.

[0062] Next, PU32 determines the presence or absence of an abnormality in the battery pack 10 by substituting the values of the variables acquired by the process of S30 into the identification model defined by the identification model data 34a (S32). Next, PU32 determines whether it is in the power limit restriction mode (S34). The power limit restriction mode is a mode that restricts the output of the battery pack 10 to the lower side.

[0063] When the PU 32 determines that the battery pack 10 is not in the power limit mode (S34: NO), the PU 32 determines whether the output of the identification model indicates that the battery pack 10 has a deterioration abnormality (S36). When the output of the identification model indicates that the battery pack 10 has a deterioration abnormality, the PU 32 switches the control mode of the battery pack 10 to the power limit mode (S40). Then, the PU 32 assigns the limit value IL to the maximum value Imax (S42).

[0064] On the other hand, if the output of the identification model does not indicate that the battery pack 10 is degraded abnormally (S36: NO), the PU 32 substitutes the normal value IH for the maximum value Imax (S38). The normal value IH is set to a value greater than the limit value IL.

[0065] When the PU 32 determines that the battery pack 10 is in the power limit mode (S34: YES), the PU 32 determines whether or not the output of the identification model indicates that the battery pack 10 is normal (S44). When the PU 32 determines that the battery pack 10 is normal (S44: YES), the PU 32 increments the counter C (S46). Then, the PU 32 determines whether or not the counter C is equal to or greater than the threshold value Cth (S48). This process is a process for determining whether or not to release the power limit mode because the battery pack 10 is considered to be normal. When the PU 32 determines that the counter C is equal to or greater than the threshold value Cth (S48: YES), the PU 32 proceeds to the process of S38. In this case, the PU 32 initializes the counter C. On the other hand, when the PU 32 determines that the counter C is less than the threshold value Cth (S48: NO) or when the process of S44 is negative, the PU 32 proceeds to the process of S42.

[0066] When the PU 32 completes the processes of S38 and S42, the PU 32 temporarily ends the series of processes shown in FIG. "Model update process" Fig. 5 shows the procedure of the process for updating the identification model data 34a and the cell model data 34b stored in the storage device 34. The process shown on the left side of Fig. 5 is realized by the PU32 repeatedly executing the program stored in the storage device 34 at a predetermined cycle, for example. The process shown on the right side of Fig. 5 is realized by the PU62 repeatedly executing the program stored in the storage device 64 at a predetermined cycle, for example. Note that the cell model data 34b stored in the storage device 34 at the time of shipment of the vehicle 2 is the same as the cell model data 64a.

[0067] In the series of processes shown in Fig. 5, the PU32 first acquires the charge / discharge current I, temperature T, terminal voltage V of the battery pack 10, and cell voltages Vc(1) to Vc(n) of the battery pack 10 (S50). Next, the PU32 calculates an estimated value Ve of the terminal voltage of the battery pack 10 based on the charge / discharge current I and temperature T as input variables using the cell model defined by the cell model data 34b (S52). Also, the PU32 calculates estimated values Vce(1) to Vce(n) of the cell voltages Vc(1) to Vc(n). Next, the PU32 updates the specified parameters and the state of charge SOC of each battery cell 12 so as to reduce the difference between the estimated value and the detected value of the voltage (S54). That is, the PU32 updates the specified parameters and the state of charge SOC of each battery cell 12 so as to reduce the difference between the cell voltages Vc(1) to Vc(n) and the estimated values Vce(1) to Vce(n), and the difference between the terminal voltage V and the estimated value Ve.

[0068] In the process of S54, as an example, the PU62 updates the maximum negative lithium ion concentration cnmax, the maximum positive lithium ion concentration cpmax, the proportional coefficient α, the intercept b, and the state of charge SOC of each battery cell 12. This process may be, for example, a process in which the PU62 searches for a solution that minimizes the difference between the estimated value and the detected value of the voltage when the values of the above specified parameters and the state of charge SOC are variously set.

[0069] Note that the process of S54 is actually executed using a plurality of sampling values at different timings between the charge / discharge current I, temperature T, terminal voltage V of the assembled battery 10, and cell voltages Vc(1) to Vc(n) of the assembled battery 10, and the corresponding estimated values.

[0070] Then, PU32 determines whether the model update has been completed (S56). For example, PU32 may determine that the model update has been completed when the absolute value of the difference between the estimated value Ve and the terminal voltage V is equal to or less than the threshold value. When PU62 determines that the model update has been completed (S56: YES), it operates the communication device 36 to transmit the updated specified parameters to the learning device 60 together with the vehicle identification symbol (S58).

[0071] On the other hand, PU62 of the learning device 60 receives the transmitted vehicle identification symbol and specified parameters (S70). Next, after updating the cell model data 64a with the received specified parameters, PU62 executes a process corresponding to the process of S14 using the updated cell model (S72). Incidentally, PU62 executes the processes of S10 and S12 using both the updated cell model data 64a and the shifted values of the specified parameter values of the cell model data 64a. That is, for example, when the value of the specified parameter of the cell model data 64a is shifted from the value of the central characteristic product but within the normal range, PU62 generates a normal battery model using that value. For example, even when the proportional coefficient α and the intercept b indicated by the cell model data 64a are shifted from the values of the central characteristic product, PU62 includes the model generated using the proportional coefficient α and the intercept b indicated by the cell model data 64a in the normal battery model when within the normal range. Also, PU62 includes, for example, a battery model generated using the proportional coefficient α and the intercept b indicated by the cell model data 64a and with the state of charge SOC varied in the abnormal battery model.

[0072] Then, similar to the process of S16, PU62 stores the data obtained by simulation as training data in the storage device 64 (S74). Then, PU72 relearns the identification model using the data stored by the process of S74 (S76). Then, by operating the communication device 66, PU62 transmits the data defining the relearned identification model to the battery ECU 30 (S78).

[0073] Note that when PU62 completes the process of S78, it temporarily ends the series of processes shown on the right side of FIG. 5. On the other hand, PU32 of the battery ECU 30 receives the data defining the identification model (S60). Then, PU32 updates the identification model data 34a stored in the storage device 34 (S62). That is, it replaces the identification model data 34a stored in the storage device 34 with the received data. Note that when PU32 completes the process of S62 and when a negative determination is made in the process of S56, it temporarily ends the series of processes shown on the left side of FIG. 5.

[0074] "Operations and Effects of the Present Embodiment" PU62 of the learning device 60 generates a model of the battery pack 10 by connecting n cell models defined by the cell model data 64a in series. In particular, PU62 represents both a normal battery pack 10 and an abnormal battery pack 10 by setting the defining parameters for the model and the SOC of each battery cell. PU62 uses the model of the battery pack 10 to calculate estimated values of the terminal voltages of each of the n battery cells 12 and the terminal voltage of the battery pack 10 based on the charge / discharge current I and the temperature T as input variables. PU62 associates a label variable for identifying whether the model of the battery pack 10 used for calculating the estimated values is a normal model or an abnormal model with the charge / discharge current I, the temperature T, the estimated values of the terminal voltages of the corresponding battery cells 12, and the estimated value of the terminal voltage of the battery pack 10. PU62 generates an identification model by using the associated data as training data.

[0075] The PU32 of the battery ECU 30 determines the presence or absence of an abnormality in the battery pack 10 by inputting the cell voltages Vc(1) to Vc(n), the terminal voltage V of the battery pack 10, the charge / discharge current I, and the temperature T into the generated identification model. When the PU32 determines that there is an abnormality, it restricts the output of the battery pack 10 in order to suppress the precipitation of lithium ions in the battery cell 12.

[0076] Here, the time required to determine the presence or absence of an abnormality using the identification model is shorter than the time required to obtain the detection values of the sensors for calculating the values of the specified parameters of the cell model. Therefore, by using the identification model, the state of the battery pack 10 can be estimated earlier than the process of updating the specified parameters of the cell model.

[0077] According to the present embodiment described above, the following operations and effects can be obtained. (1) The PU62 of the learning device 60 independently sets the values of the specified parameters that define the respective cell models of the battery cells 12(1) to 12(n). Thereby, the degree of freedom in expressing the state of the battery pack 10 can be increased as compared with the case where the values of the specified parameters that define the respective cell models of the battery cells 12(1) to 12(n) are common.

[0078] (2) The PU62 of the learning device 60 uses, as training data, not only the simulation results using the battery model but also the measured values of the actual behavior of the actual battery pack 10 for the learning of the identification model. Thereby, the accuracy of the output of the identification model can be improved.

[0079] (3) Using the specified parameters of the cell model calculated by the PU32 of the battery ECU 30, the PU62 of the learning device 60 generated training data for re-learning. Then, the PU62 learned an identification model specific to the battery pack 10 mounted on one vehicle. Thereby, as compared with the identification model generated by the process of FIG. 3, a model capable of highly accurately determining the presence or absence of an abnormality in the battery pack 10 mounted on one vehicle can be realized.

[0080] That is, for example, a battery model representing the variation abnormality of the state of charge (SOC) of battery cells 12(1) to 12(n) may have values of specified parameters different from those of the specified parameters that accurately represent the battery pack 10 to be estimated. In that case, the accuracy of determining the presence or absence of the variation abnormality of the state of charge (SOC) of battery cells 12(1) to 12(n) may decrease due to the identification model generated by the process of FIG. 3. In contrast, in the present embodiment, the PU 62 of the learning device 60 re-learns the identification model using the values of the specified parameters that accurately represent the battery pack 10 to be estimated. Therefore, it is possible to accurately determine the presence or absence of an abnormality in the battery pack 10 to be estimated.

[0081] (4) The learning device 60 executed re-learning of the identification model. As a result, the computational load on the battery ECU 30 can be reduced as compared with the case where the battery ECU 30 executes re-learning of the identification model.

[0082] <Corresponding relationship> The correspondence between the matters in the above-described embodiment and the matters described in the column of "Means for Solving the Problems" is as follows. Below, the correspondence is shown for each number of the solution means described in the column of "Means for Solving the Problems". [1, 2, 3] The model definition process corresponds to the processes of S10 and S12. The training data generation process corresponds to the process of S14. The learning process corresponds to the process of S20. The input / output model corresponds to a model in which cell models are connected in series. [4] The training data corresponds to the data obtained by the process of S18. [5] The individual data calculation process corresponds to the processes of S52 to S56. The training data regeneration process corresponds to the process of S72. The relearning process corresponds to the process of S76. [6] The input variable acquisition process corresponds to the process of S30. The state estimation process corresponds to the process of S32. [7, 8] The model definition process corresponds to the processes of S10 and S12. The training data generation process corresponds to the process of S14. The learning process corresponds to the process of S20. The input / output model corresponds to a model in which cell models are connected in series. The input variable acquisition process corresponds to the process of S30. The state estimation process corresponds to the process of S32. [9, 10] The individual data calculation process corresponds to the processes of S52 to S56. The specified parameter transmission process corresponds to the process of S58. The state estimation model reception process corresponds to the process of S60. The specified parameter reception process corresponds to the process of S70. The training data regeneration process corresponds to the process of S72. The relearning process corresponds to the process of S76. The state estimation model transmission process corresponds to the process of S78.

[0083] <Other Embodiments> Note that this embodiment can be implemented with the following modifications. This embodiment and the following modification examples can be implemented in combination with each other within a technically non-conflicting range.

[0084] "Regarding Training Data" · The temperature included in the training data is not limited to the temperature T of the battery pack 10. For example, it may be a temperature independently set for each of the battery cells 12(1) to 12(n). When the training data includes data having different temperatures among the battery cells 12(1) to 12(n), the situation that can actually occur in the battery pack 10 can be expressed with higher accuracy.

[0085] "Regarding the state estimation model" · It is not essential that the state estimation model is an identification model. For example, the state estimation model may be a regression model that outputs the full charge capacity of the battery cells 12(1) to 12(n). In other words, the state to be estimated by the state estimation model may be the full charge capacity. Also, for example, the state estimation model may be a regression model that outputs a variable value indicating the degree of variation in the state of charge (SOC) of the battery cells 12(1) to 12(n). Here, the variable value indicating the degree of variation may be, for example, the maximum value of the absolute value of the difference in the state of charge (SOC) of the battery cell 12 with respect to the average value of the state of charge (SOC) of the battery cells 12(1) to 12(n).

[0086] "Regarding the battery" · It is not essential that the battery to be the object of state estimation by the state estimation model is a battery pack.

[0087] "Regarding the input / output model" · The temperature T input to the cell model is not limited to the temperature T of the battery pack 10. For example, if it is possible to obtain different temperatures for each of the battery cells 12(1) to 12(n), it may be the unique temperature of the target battery cell 12.

[0088] · The cell model in which current is included in the input is not limited to the model illustrated in FIG. 2. · The model in which either one of the current and voltage of the battery is included in the input and the other is included in the output is not limited to the model in which current is included in the input. For example, it may be a model in which voltage is included in the input.

[0089] "Regarding the state estimation system" ·All of the processes of S50 to S56, S62, and S72 to S76 in FIG. 5 may be executed by the battery ECU 30.

[0090] ·All of the processes of S52 to S56 and S72 to S78 in FIG. 5 may be executed by the learning device 60. In that case, the PU 32 of the battery ECU 30 may execute the process of S50, the process of transmitting the data acquired by the process of S50 to the learning device 60, and the processes of S60 and S62.

[0091] ·It is not essential for the state estimation system to include the battery ECU 30 mounted on the vehicle and the learning device 60 capable of communicating with a plurality of vehicles. For example, the state estimation system may constitute a control device provided in a power generation facility equipped with a battery.

[0092] "Regarding the state estimation device" ·The state estimation device is not limited to one that executes software processing. For example, it may include a dedicated hardware circuit such as an ASIC that executes at least a part of the processes executed in the above-described embodiment. That is, the state estimation process may include a processing circuit having any of the following configurations (a) to (c). (a) A processing circuit including a processing device that executes all of the above processes according to a program and a program storage device such as a storage device that stores the program. (b) A processing circuit including a processing device and a program storage device that execute a part of the above processes according to a program, and a dedicated hardware circuit that executes the remaining processes. (c) A processing circuit including a dedicated hardware circuit that executes all of the above processes. Here, there may be a plurality of software execution devices including a processing device and a program storage device. Also, there may be a plurality of dedicated hardware circuits.

[0093] "Regarding the learning device" ·The learning device is not limited to those that execute software processing. For example, it may include a dedicated hardware circuit such as an ASIC that executes at least a part of the processing executed in the above-described embodiment. That is, the learning process may include a processing circuit having any of the following configurations (a) to (c). (a) A processing circuit including a processing device that executes all of the above processing according to a program, and a program storage device such as a storage device that stores the program. (b) A processing circuit including a processing device and a program storage device that execute a part of the above processing according to a program, and a dedicated hardware circuit that executes the remaining processing. (c) A processing circuit including a dedicated hardware circuit that executes all of the above processing. Here, there may be a plurality of software execution devices including a processing device and a program storage device. Also, there may be a plurality of dedicated hardware circuits.

Description of Signs

[0094] 2…Vehicle 4…Motor generator 6…Power conversion circuit 10…Battery pack 12…Battery cell 22…Current sensor 50…Network

Claims

1. A method for training a state estimation model for estimating the state of a battery, comprising: executing a model definition process, a training data generation process, and a training process; the model definition process is a process of setting, as values representing respective states of different batteries, definition parameters that are parameters for defining an input-output model; the input-output model is a model in which either one of the current and voltage of the battery is included in the input and the other is included in the output; the training data generation process is a process of generating training data for the state estimation model; the training data includes data of input and output obtained by simulation using the input-output model and data indicating the state of the battery represented by the value of the definition parameter of the input-output model; the training process is a process of training the state estimation model with the training data; the input variables of the state estimation model include the input and output of the input-output model; the output of the state estimation model is a method for training a state estimation model that indicates the state of the battery.

2. The battery is a battery pack in which a plurality of battery cells are connected in series, the input-output model is a model in which a plurality of cell models are connected in series, the cell model is a model in which either one of the battery cells is included in the input and the other is included in the output; The method for training a state estimation model according to claim 1, wherein the model definition process includes a process of expressing the state of the battery pack by determining the value of the definition parameter for each cell model.

3. The method for training a state estimation model according to claim 1, wherein the state estimation model is a discrimination model for estimating the presence or absence of progress of deterioration of the battery.

4. The training data used for training the state estimation model by the training process includes, in addition to the training data generated by the training data generation process, data obtained when an actual battery is charged and discharged and data indicating the state of the battery. The method for training a state estimation model according to claim 1, which includes a set of data.

5. comprising steps of executing an individual data calculation process, a training data regeneration process, and a retraining process; the individual data calculation process is a process of calculating the value of the definition parameter by measuring the voltage and current of one battery. The training data regeneration process includes a simulation using the input-output model defined by the value of the specified parameter calculated by the individual data calculation process, to generate data of the input and output of the input-output model and data indicating the state of the battery represented by the value of the specified parameter of the input-output model. The relearning process is a process of relearning the state estimation model using the training data generated by the training data regeneration process. The method for learning a state estimation model according to claim 1.

6. A method for manufacturing a state estimation device for estimating the state of a battery, The state estimation device is configured to execute an input variable acquisition process and a state estimation process. The input variable acquisition process is a process of acquiring the value of the input variable of the state estimation model in the method for learning a state estimation model according to claim 1. The state estimation process is a process of estimating the state of the battery by inputting the value of the input variable acquired by the input variable acquisition process into the state estimation model. A method for manufacturing a state estimation device having each step in the method for learning a state estimation model according to claim 1.

7. A state estimation system for estimating the state of a battery, It is configured to execute a model definition process, a training data generation process, a learning process, an input variable acquisition process, and a state estimation process. The model definition process is a process of setting a specified parameter, which is a parameter for defining an input-output model, to a value representing each of different states of the battery. The input-output model is a model in which either one of the current and voltage of the battery is included in the input and the other is included in the output. The training data generation process is a process of generating training data for the state estimation model. The training data includes data of the input and output obtained by simulation using the input-output model and data indicating the state of the battery represented by the value of the specified parameter of the input-output model. The learning process is a process of learning the state estimation model with the training data. The input variables of the state estimation model include the input and output of the input-output model. The output of the state estimation model indicates the state of the battery. The input variable acquisition process is a process of acquiring the value of the input variable of the state estimation model. The state estimation system is a system that estimates the state of the battery by inputting the values obtained by the input variable acquisition process into the state estimation model.

8. Comprising a state estimation device and a learning device, The state estimation device is configured to execute the input variable acquisition process and the state estimation process, The learning device is configured to execute the model definition process, the training data generation process, and the learning process, according to the state estimation system described in Claim 7.

9. The state estimation device is configured to execute an individual data calculation process, a specified parameter transmission process, and a state estimation model reception process, The learning device is configured to execute a specified parameter reception process, a training data regeneration process, a relearning process, and a state estimation model transmission process, The individual data calculation process is a process of calculating the values of the specified parameters by measuring the voltage and current of one battery, The specified parameter transmission process is a process of transmitting the specified parameters calculated by the individual data calculation process to the learning device, The specified parameter reception process is a process of receiving the specified parameters transmitted by the specified parameter transmission process, The training data regeneration process is a process of generating data of the input and output of the input-output model and data indicating the state of the battery represented by the values of the specified parameters of the input-output model by simulation using the input-output model defined by the values of the specified parameters received by the specified parameter reception process, The relearning process is a process of relearning the state estimation model using the training data generated by the training data regeneration process, The state estimation model transmission process is a process of transmitting the state estimation model relearned by the relearning process to the state estimation device, The state estimation model reception process is a process of receiving the state estimation model transmitted by the state estimation model transmission process, according to the state estimation system described in Claim 8.

10. The state estimation device in the state estimation system described in Claim 9.

Images