Battery degradation state estimation method and battery degradation state estimation device

Abstract

A controller (23) constructs a model with measurement information including AC resistance values measured for training and usage history information acquired for training as explanatory variables, and a deterioration state measured for training as an objective variable. A controller (43) uses the constructed model to perform estimation with measurement information including AC resistance values measured for estimation and usage history information acquired for estimation as explanatory variables, and the deterioration state of a battery (42) as an objective variable.

Description

【Technical Field】 【0001】 The present invention relates to a method for estimating the state of battery deterioration and an apparatus for estimating the state of battery deterioration. 【Background Art】 【0002】 In Patent Document 1, it is proposed to previously store the correspondence relationship between the temperature history, input / output resistance, and full charge capacity of a secondary battery, and estimate the full charge capacity of the secondary battery from the temperature history and input / output resistance of the secondary battery. 【Prior Art Documents】 【Patent Documents】 【0003】 【Patent Document 1】 Japanese Unexamined Patent Application Publication No. 2018-132369 【Summary of the Invention】 【Problems to be Solved by the Invention】 【0004】 Since it is not possible to know how the battery has been used only from the temperature history, it is difficult to accurately estimate the state of battery deterioration by considering only the temperature history. An object of the present invention is to improve the accuracy of estimating the state of deterioration in a battery. 【Means for Solving the Problems】 【0005】 According to one aspect of the present invention, the learning controller measures the AC resistance of a battery to be learned mounted on a vehicle using electrochemical impedance spectroscopy and performs a process to acquire measurement information including at least the real axis component of the AC resistance measured for learning. The learning controller performs a process to acquire usage history information of the battery to be learned for learning. The learning controller performs a process to measure the degradation state of the battery to be learned for learning. The learning controller performs a process to construct a model using the measurement information measured for learning and the usage history information acquired for learning as explanatory variables, and the degradation state measured for learning as the objective variable. The estimation controller measures the AC resistance of a battery to be estimated using electrochemical impedance spectroscopy and performs a process to acquire measurement information including at least the real axis component of the AC resistance measured for estimation. The estimation controller performs a process to acquire usage history information of a battery to be estimated for estimation. The estimation controller uses the model and performs a process to estimate the degradation state of the battery to be estimated as the objective variable, using the measurement information measured for estimation and the usage history information acquired for estimation as explanatory variables. [Effects of the Invention] 【0006】 According to the present invention, by considering the battery's usage history information, the accuracy of estimating the degradation state of the battery can be improved. [Brief explanation of the drawing] 【0007】 [Figure 1] This is a block diagram of a battery degradation state estimation device. [Figure 2] This is a flowchart of the learning process. [Figure 3] This is a Nyquist plot of AC resistance values. [Figure 4] This is a schematic diagram of e-dimensional data. [Figure 5] This is a schematic diagram of m-dimensional data. [Figure 6] This is a schematic diagram of the design matrix. [Figure 7] This is a schematic diagram of N-dimensional data. [Figure 8] This is a flowchart showing the estimation process. [Figure 9] This figure shows the second embodiment. [Figure 10] This figure shows the third embodiment. [Modes for carrying out the invention] 【0008】 Embodiments of the present invention will be described below with reference to the drawings. Note that the drawings are schematic and may differ from actual examples. Furthermore, the following embodiments are illustrative examples of devices and methods for realizing the technical concept of the present invention, and do not limit the configuration to those described below. In other words, the technical concept of the present invention can be modified in various ways within the technical scope described in the claims. 【0009】 《First Embodiment》 "composition" Figure 1 is a block diagram of a battery degradation state estimation device. The battery degradation state estimation device 11 comprises a vehicle to be learned 21 and a vehicle to be estimated 41. Vehicle 21 is an electric vehicle or hybrid vehicle that can be driven by a motor and is equipped with a battery 22 (battery to be learned) and a controller 23 (learning controller). The battery 22 is a lithium-ion battery used for vehicle propulsion, but can also be a sodium-ion battery or a total solid Batteries will also work. 【0010】 The battery 22 is equipped with an EIS sensor 31, a current sensor 32, a voltage sensor 33, and a temperature sensor 34. The EIS sensor 31 measures the AC resistance Z of the battery 22 by electrochemical impedance spectroscopy (EIS). The current sensor 32 measures the current I of the battery 22. The voltage sensor 33 measures the voltage E of the battery 22. The temperature sensor 34 measures the temperature Tb of the battery 22. The vehicle 21 is equipped with an outside air temperature sensor 35, a positioning device 36, and a charge / discharge controller 37. The outside air temperature sensor 35 measures the outside air temperature To. The positioning device 36 acquires the location information of the vehicle 21 via GPS. When the battery 22 is charged, the charge / discharge controller 37 acquires charge type information, such as whether it is a normal charge or a fast charge, and when the battery 22 is discharged, it acquires discharge type information, such as whether it is discharged due to driving or discharged due to other reasons, such as supplying power to a home or smart grid. The charge type information and discharge type information can be acquired via the vehicle 21's controller 23, which will be described later, or via a charger management device or a power supply destination management device, etc. 【0011】 The controller 23 is composed of, for example, a microcomputer and includes a measurement information acquisition unit 24, a history information acquisition unit 25, a degradation state measurement unit 26, and a model construction unit 27. The measurement information acquisition unit 24 acquires measurement information including at least the real axis component Zr of the AC resistance value Z acquired by the EIS sensor 31. The history information acquisition unit 25 acquires usage history information of the battery 22. The degradation state measurement unit 26 measures the degradation state (SOH: States Of Health) of the battery 22 using an existing method. The model construction unit 27 constructs a model for each battery cell type with the measurement information measured for training and the usage history information acquired for training as explanatory variables, and the degradation state SOH measured for training as the objective variable. 【0012】 Vehicle 41 includes a battery 42 (estimated target battery) and a controller 43 (estimating controller). The battery 42 is of the same battery cell type as the battery 22 mounted on the vehicle 21 to be learned. The battery 42 is provided with an EIS sensor 51, a current sensor 52, a voltage sensor 53, and a temperature sensor 54. The vehicle 41 is provided with an outside air temperature sensor 55, a positioning device 56, and a charge / discharge controller 57. The controller 43 includes a measurement value information acquisition unit 44, a history information acquisition unit 45, a model storage unit 46, and a deterioration state estimation unit 47. For parts common to the configuration of the vehicle 21, detailed description will be omitted. The model storage unit 46 stores the model constructed by the model construction unit 27. The deterioration state estimation unit 47 uses the model, and estimates the state of health (SOH) of the battery 42 with the measured measurement value information measured for estimation and the usage history information acquired for estimation as explanatory variables and the deterioration state of the battery 42 as the objective variable. 【0013】 Next, the learning process executed by the controller 23 of the vehicle 21 to be learned will be described. This process is executed at regular intervals, but since the deterioration of the battery 22 does not progress daily, for example, it may be executed once a week or once a month, but it is of course not limited to these. FIG. 2 is a flowchart showing the learning process. First, in step S101, the AC resistance value Z, which is the AC impedance of the battery 22, is measured by the EIS sensor 31. The measurement of the AC resistance value Z can be carried out at any timing, but in order to improve the measurement accuracy, it is desirable to carry it out in a quiescent state where neither running discharge nor charging is performed and a predetermined time has elapsed after parking. FIG. 3 is a Nyquist plot of the AC resistance value Z. Here, with the real-axis component Zr of the AC resistance value Z as the horizontal axis and the imaginary-axis component Zx of the AC resistance value Z as the vertical axis, the AC resistance value Z corresponding to the frequency fa is plotted. The AC resistance value Z to be measured is in the frequency band of 100 Hz or more where there is no significant variation in the real-axis component Zr. 【0014】 Return to the description of FIG. 2. In the subsequent step S102, e-dimensional data (first-dimensional data) having e feature amounts is generated from the measurement value information measured for learning. e is a natural number of 1 or more. The measurement value information includes at least the real-axis component Zr of the AC resistance value Z. The measurement value information may include, as other feature amounts, any one of the imaginary-axis component Zx of the AC resistance value Z, the difference ΔZ between the DC resistance value DCR and the real-axis component Zr, the state of charge SOC (State Of Charge), and the open-circuit voltage OCV. FIG. 4 is a schematic diagram of the e-dimensional data. One feature amount is schematically represented by one square, and a plurality of feature amounts are drawn side by side in a row. Here, as an example, e-dimensional data having three feature amounts is used. 【0015】 Returning to the explanation of Figure 2, in the following step S103, usage history information of the battery 22 is obtained. The usage history information includes at least one of the following: the maximum battery temperature TbMAX during charging and discharging, the time spent in the battery temperature zone Tstemp during the parking period, the time spent in the charge state zone Tssoc during the parking period, the charge state SOC at the start of charging and discharging, the battery temperature Tb at the start of charging and discharging, the charge / discharge current Qc, and the charge / discharge type. The usage history information is maintained as time-series data with discharge and charging each as one event, and one or more feature quantities for each event. Discharge includes not only discharge due to driving, but also discharge due to power supply to homes and smart grids. Depending on the mounting position of the module, there may be variations in cooling performance, and degradation progresses when exposed to high-temperature environments. Therefore, for battery temperature, statistics such as maximum, mean, and variance related to the module mounting position and ambient temperature To, as well as regional classification using mean latitude information and the mode of latitude and longitude from the vehicle 21's location information, may be added. The battery temperature zone is, for example, a temperature zone divided into low temperature, medium temperature, and high temperature. The charge / discharge current Qc is the total value during an event. A higher frequency of power supply indicates a higher charge / discharge cycle frequency. Similarly, a high variance in the charge state (SOC) at the start of charging indicates a wide range of charge state (SOC) used (swing range), which also indicates a high charge / discharge cycle frequency. This charge / discharge history is also an indicator that influences the degradation state (SOH). The usage history information may also include the proportion of the discharge current due to power supply to the total charge current, as well as cell type information for the battery 22 being studied. 【0016】 In the following step S104, m-dimensional data (second-dimensional data) with m features is generated from the usage history information acquired for training. m is a natural number greater than or equal to 1. Specifically, m-dimensional data is generated by performing statistical processing on the usage history information held as time-series data. For example, statistics including sum, maximum, minimum, mean, quartiles, mode, and variance are used. Feature selection methods common in model training are used to select features of high importance, such as using a decision tree model or PFI, or selecting features using a Lasso regression model. Figure 5 is a schematic diagram of m-dimensional data. Each feature is schematically represented by a single square, and multiple features are plotted in a single row. Here, as an example, we use m-dimensional data with five features. 【0017】 Returning to the explanation of Figure 2, in the following step S105, a design matrix is ​​generated by collecting N e+m-dimensional data (third-dimensional data) which are combinations of the e-dimensional data and the m-dimensional data generated for training. N is a natural number greater than or equal to e+m, and is the number of sample data required for machine learning. In order to improve training accuracy, it is necessary to set N to a large value. The design matrix may be subjected to standardization operations such as setting the mean to zero and the variance to one, setting the median to zero and the interquartile range to one, and restricting the range of variables to a specific interval. It may also be subjected to processing such as nonlinear transformation by logarithmization. Figure 6 is a schematic diagram of the design matrix. By arranging the e-dimensional data generated for training and the m-dimensional data generated for training into a single row to form e+m-dimensional data, and then arranging this e+m-dimensional data into a single column, a design matrix with N rows and e+m columns is generated. 【0018】 Returning to the explanation of Figure 2, in the following step S106, the state of health (SOH) of the battery 22 is measured for learning purposes. Here, as an existing method, the state of health (SOH) of the battery 22 is measured, for example, by measuring the cumulative discharge amount from a fully charged state. In the following step S107, N-dimensional data (fourth-dimensional data) is generated from the N degraded states (SOH) measured for training. The number of N-dimensional data corresponds to the number of rows in the design matrix. Figure 7 is a schematic diagram of N-dimensional data. Each degradation state of Health (SOH) is schematically represented by a single square, and multiple degradation states of SOH are depicted arranged in a line. 【0019】 Return to the description of FIG. 2. In the subsequent step S108, a model is constructed and stored using the planned matrix generated for learning as an explanatory variable and the N-dimensional data generated for learning as an objective variable. The model is constructed for each battery cell type. Usually, since the degradation state SOH is expressed in the range of SOH = {SOH|0 < SOH < 1}, it may be configured to learn a regression model with z obtained by performing the logit transformation shown in Equation 1 below, that is, z = {z|-∞ < z < ∞} as the objective variable. Thereby, it is possible to avoid the estimated value of the degradation state SOH using the model from becoming less than 0 or exceeding 1. Here, a Gaussian process regression model using a kernel function is adopted, but it is not limited thereto. As the machine learning algorithm, other algorithms such as lasso regression, elastic net regression, support vector regression, random forest, and neural network may also be adopted. 【Equation】 【0020】 Note that when the number of e + m-dimensional data and the number of N-dimensional data are less than N, the process returns to the main program as it is. Therefore, until the number of e + m-dimensional data and the number of N-dimensional data reach N, it becomes a loop process that repeats the processing of steps S101 to S107. When the number of e + m-dimensional data and the number of N-dimensional data reach N, an initial model is constructed. Thereafter, the learned model is updated by increasing or updating the number of e + m-dimensional data and N-dimensional data. The above is the learning process executed by the controller 23. Among them, the process of step S101 corresponds to the measurement value information acquisition unit 24, and the process of step S103 corresponds to the history information acquisition unit 25. Also, the process of step S106 corresponds to the degradation state measurement unit 26, and the process of step S108 corresponds to the process of the model construction unit 27. 【0021】 Next, the estimation process executed by the controller 43 of the vehicle 41 to be estimated will be described. This process is executed at regular intervals. However, since the deterioration of the battery 42 does not progress daily, it may be executed, for example, once a week or once a month, but it is of course not limited to these frequencies. FIG. 8 is a flowchart showing the estimation process. First, in step S111, the AC resistance value Z, which is the AC impedance of the battery 42, is measured by the EIS sensor 51. The concept is the same as the process of step S101 during learning. In the subsequent step S112, e-dimensional data (first-dimensional data) having e feature amounts is generated from the measurement value information measured for estimation. The concept is the same as the process of step S102 during learning. 【0022】 In the subsequent step S113, the usage history information of the battery 42 is acquired. The concept is the same as the process of step S103 during learning. In the subsequent step S114, m-dimensional data (second-dimensional data) having m feature amounts is generated from the usage history information acquired for estimation. The concept is the same as the process of step S104 during learning. In the subsequent step S115, the state of health SOH of the battery 42 is estimated using the model stored in the model storage unit 46. That is, as shown in Equation 2 below, e+m-dimensional data (third-dimensional data) obtained by combining the e-dimensional data generated for estimation and the m-dimensional data generated for estimation is used as an explanatory variable, and the state of health SOH of the battery 42 is estimated as a target variable. 【Equation】 【0023】 When logit transformation is performed during learning, for the estimated value from the learned regression model, by performing the inverse logit transformation shown in Equation 3 below, the state of health SOH expressed in the range of SOH={SOH|0<SOH<1} can be obtained. 【Equation】 The above describes the estimation process performed by the controller 43. Of these, the process in step S111 corresponds to the measurement value information acquisition unit 44, the process in step S113 corresponds to the history information acquisition unit 45, and the process in step S115 corresponds to the deterioration state estimation unit 47. 【0024】 "effect" Next, the main effects of the first embodiment will be described. (1) Controller 23 measures the AC resistance Z of the battery 22 used for vehicle propulsion using electrochemical impedance spectroscopy and performs a process to acquire measurement information including at least the real-axis component Zr of the AC resistance Z measured for learning. Controller 23 performs a process to acquire usage history information of the battery 22 for learning and a process to measure the state of deterioration (SOH) of the battery 22 for learning. Controller 23 performs a process to construct a model in which the measurement information measured for learning and the usage history information acquired for learning are explanatory variables and the state of deterioration (SOH) measured for learning is the objective variable. The degradation mode differs depending on how the battery 22 is used, but by performing machine learning considering the usage history information of the battery 22, a model can be constructed that improves the accuracy of estimating the state of deterioration (SOH). Controller 43 measures the AC resistance Z of the battery 42 used for vehicle propulsion using electrochemical impedance spectroscopy and performs a process to acquire measurement information including at least the real-axis component Zr of the AC resistance Z measured for estimation. Controller 43 performs a process to acquire usage history information of the battery 42 for estimation. The controller 43 uses a model to perform a process that estimates the state of health (SOH) of the battery 42, using the measured value information and usage history information acquired for estimation as explanatory variables, and the state of health (SOH) of the battery 42 as the objective variable. Although the degradation mode differs depending on how the battery 42 is used, the accuracy of estimating the state of health (SOH) of the battery 42 can be improved by considering the usage history information of the battery 42. 【0025】 (2) The measured value information includes, in addition to the real axis component Zr, at least one of the following: the imaginary axis component Zx of the AC resistance value Z, the difference ΔZ between the DC resistance value DCR and the real axis component Zr, the charge state (SOC), and the open-circuit voltage (OCV). By considering features that can represent changes in the Nyquist plot in addition to the real axis component Zr, the estimation accuracy of the degradation state (SOH) can be further improved. Furthermore, a value equivalent to the electrode resistance can be simulated as the interfacial resistance. In addition, the effect of electron and ion conduction due to changes in the positive and negative electrode volumes can be considered. (3) Usage history information includes the maximum battery temperature Tb during charging and discharging. MAX The data includes at least one of the following: Tstemp (time spent in the battery temperature zone during parking), Tssoc (time spent in the charge state zone during parking), SOC (charge state at the start of charging / discharging), Tb (battery temperature at the start of charging / discharging), Qc (charging / discharging current), and Qc (charging / discharging type). This allows for consideration of factors affecting the degradation state (SOH), reducing the computational load using more limited data, and improving the accuracy of SOH estimation. (4) The AC resistance value Z is in the frequency band of 100 Hz or higher. This allows for the measurement of the electrolyte resistance component of the AC resistance value Z that has a high correlation with the degradation state SOH. 【0026】 (5) The controller 23 performs the following processes: generating e-dimensional data with e features from measurement data measured for training, and generating m-dimensional data with m features from usage history information acquired for training. The controller 23 then performs the following processes: generating a design matrix by collecting N e+m-dimensional data, which are combinations of the e-dimensional data and the m-dimensional data generated for training, so that there are at least e+m features. The controller 23 then processes the N degradation states measured for training. SOHThe controller 23 executes a process to generate N-dimensional data. The controller 23 constructs a model using the design matrix generated for training and the N-dimensional data generated for training. This allows for the construction of a model that improves the accuracy of estimating the degradation state (SOH) by performing machine learning while considering the usage history information of the battery 22. The controller 43 executes a process to generate e-dimensional data with e features from the measured value information measured for estimation, and a process to generate m-dimensional data with m features from the usage history information acquired for estimation. The controller 43 uses the model to estimate the degradation state (SOH) from the e+m-dimensional data, which is a combination of the e-dimensional data generated for estimation and the m-dimensional data generated for estimation. This allows for the improvement of the accuracy of estimating the degradation state (SOH) in the battery 42 by considering the usage history information of the battery 42. 【0027】 The controller 23 constructs a model for each battery cell type. This is because the impact of the degradation state (SOH) on various features differs for each battery cell type. Therefore, by constructing a model for each battery cell type, it is possible to construct a model that improves estimation accuracy. Controller 23 is implemented in the vehicle 21 being trained, and controller 43 is implemented in the vehicle 41 being estimated. This allows for direct access to primary data, enabling calculations using high temporal resolution and information resolution data over the vehicle's service life. Consequently, highly accurate estimation of the state of deterioration (SOH) becomes possible. 【0028】 Next, we will explain the comparative examples. Here, we use a configuration that estimates the state of battery degradation (SOH) by considering the AC resistance value Z measured by an EIS sensor and the battery's temperature history as a comparative example. Even when considering the temperature history, if the battery is used in a way that involves frequent charging and discharging, the correlation between the AC resistance value Z and the degradation state (SOH) will change. This is related to the occurrence of lithium electrodeposition, the blocking of ion and electron conduction paths leading to the isolation of active material, and the deactivation of the reaction surface. Therefore, since the temperature history alone does not reveal how the battery was used, considering only the temperature history does not allow us to determine the state of battery degradation. SOH Accurately estimating this is difficult. Furthermore, even when correcting using a reference calibration curve based on AC resistance Z, battery degradation is influenced by a complex interplay of multiple factors, including charge state (SOC), temperature history, and load cycles, making a simple regression model insufficient. 【0029】 《Second Embodiment》 "composition" The second embodiment involves implementing the controller 43, which estimates the state of deterioration (SOH), on an external server. Since the other configurations are the same as those of the first embodiment described above, detailed explanations of the common parts will be omitted. Figure 9 shows a second embodiment. Here, an external server 61 is provided, and a controller 43 is implemented on this external server 61. The vehicle 41 is equipped with a communication device 62, which communicates with the external server 61 via the internet. The external server 61 is equipped with a communication device 63, which communicates with the vehicle 41 via the internet. The controller 43 receives various information measured by the vehicle 41, estimates the state of deterioration (SOH), and transmits the estimation result to the vehicle 41. 【0030】 "effect" Next, the main effects of the second embodiment will be described. (1) The controller 43 is implemented on an external server 61 that can communicate with the vehicle 41. This reduces the computational load and data volume on the vehicle 41 while improving the estimation accuracy of the deterioration state of health (SOH). In addition, since the updated model can be used in a cloud environment, the estimation accuracy of the deterioration state of health (SOH) can be improved. Other effects and benefits resulting from the common configuration are the same as those of the first embodiment described above. 【0031】 Variant form In the second embodiment, a configuration was described in which only the controller 43 that estimates the deterioration state of health (SOH) is implemented on an external server, but the system is not limited to this. That is, the controller 23 that builds the model may also be implemented on an external server that can communicate with the vehicle 21. This reduces the computational load and data capacity performed on the vehicle 21 and makes it possible to build a model that improves the estimation accuracy of the deterioration state of health (SOH). Furthermore, since the design matrix and N-dimensional data generated for training can be updated in a cloud environment, it is possible to build a model that improves the estimation accuracy of the deterioration state of health (SOH). In addition, since measurement information and usage history information provided from multiple vehicles 21 can be referenced in a unified manner, it is possible to build a model that improves the estimation accuracy of the deterioration state of health (SOH). 【0032】 Third Embodiment "composition" The third embodiment involves implementing the estimation controller 43 on an external server to estimate the state of health (SOH) of the battery 42 removed from the vehicle 41. Since the other configurations are the same as those of the first embodiment described above, detailed explanations of the common parts will be omitted. Figure 10 shows a third embodiment. Here, an external server 71 is provided, and a controller 43 is implemented on this external server 71. An EIS sensor 72 is attached to the battery 42 removed from the vehicle 41. The measurement value information acquisition unit 44 acquires measurement value information including at least the real axis component Zr of the AC resistance value Z acquired by the EIS sensor 72. The controller 43 includes a vehicle data storage unit 73. The vehicle data storage unit 73 stores the usage of the battery 22. history It stores information. The history information acquisition unit 45 uses the vehicle data storage unit 73 to determine the usage of the battery 22. history Information is acquired. Although the AC resistance value Z is measured using the EIS sensor 72, this is not the only method. That is, if the AC resistance value Z last measured on the vehicle 41 can be obtained from the vehicle data storage unit 73, it can be used. Also, for used batteries that do not have the EIS sensor 72 attached, the resistance can of course be measured using an EIS measuring device installed offline (not mounted in a vehicle, so-called off-board). 【0033】 "effect" Next, the main effects of the third embodiment will be described. (1) The controller 43 estimates the State of Health (SOH) of the battery 42 removed from the vehicle 41. This improves the accuracy of estimating the SOH of the used battery 42. As a result, the reuse of the battery 42 can be promoted. Other effects and benefits resulting from the common configuration are the same as those of the first embodiment described above. 【0034】 Although the above description has been made with reference to a limited number of embodiments, the scope of the rights is not limited to those embodiments, and modifications of the embodiments based on the above disclosure will be obvious to those skilled in the art. [Explanation of symbols] 【0035】 11...Battery degradation state estimation device, 21...Vehicle, 22...Battery, 23...Controller, 24...Measurement value information acquisition unit, 25...History information acquisition unit, 26...Degradation state measurement unit, 27...Model construction unit, 31...EIS sensor, 32...Current sensor, 33...Voltage sensor, 34...Temperature sensor, 35...Ambient temperature sensor, 36...Positioning device, 37...Charge / discharge controller, 41...Vehicle, 42...Battery, 43...Controller, 44...Measurement value information acquisition unit, 45...History information acquisition unit, 46...Model storage unit, 47...Degradation state estimation unit, 51...EIS sensor, 52...Current sensor, 53...Voltage sensor, 54...Temperature sensor, 55...Ambient temperature sensor, 56...Positioning device, 57...Charge / discharge controller, 61...External server, 62...Communication device, 63...Communication device, 71...External server, 72...EIS sensor, 73...Vehicle data storage unit

Claims

[Claim 1] A process to measure the AC resistance of a battery used for vehicle propulsion by electrochemical impedance spectroscopy, and to obtain measurement information including at least the real axis component of the AC resistance measured for learning, A process to acquire usage history information of the battery to be studied for learning purposes, A process for measuring the degradation state of the battery to be studied for learning purposes, The learning controller is instructed to perform the following process: construct a model using the measured value information and the usage history information acquired for learning as explanatory variables, and the degradation state measured for learning as the target variable. A process to measure the AC resistance value of the battery to be estimated for vehicle propulsion using electrochemical impedance spectroscopy, and to obtain measurement information including at least the real axis component of the AC resistance value measured for estimation, A process to acquire usage history information of the battery to be estimated for estimation, Using the aforementioned model, the estimation controller is instructed to perform a process that estimates the degradation state of the battery to be estimated, using the measured value information and the usage history information acquired for estimation as explanatory variables, and the degradation state of the battery to be estimated as the objective variable. The aforementioned usage history information is characterized by being time-series data in which battery discharge and charging are each treated as separate events, and each event has one or more feature quantities. [Claim 2] The battery degradation state estimation method according to claim 1, wherein the measured value information includes, in addition to the real axis component, at least one of the imaginary axis component of the AC resistance value, the difference between the DC resistance value and the real axis component, the charge state, and the open-circuit voltage. [Claim 3] The battery degradation state estimation method according to claim 1, wherein the usage history information includes at least one of the following: maximum battery temperature during charging and discharging, duration of stay in the battery temperature zone during the parking period, duration of stay in the charged state zone during the parking period, charge state at the start of charging and discharging, battery temperature at the start of charging and discharging, charge and discharge current amount, and charge and discharge type. [Claim 4] The battery degradation state estimation method according to claim 1, wherein the AC resistance value is in the frequency band of 100 Hz or higher. [Claim 5] The aforementioned learning controller is: A process for generating first-dimensional data having predetermined quantities of features from the measured value information measured for learning, A process for generating second-dimensional data having a predetermined number of feature quantities from the aforementioned usage history information acquired for training, A process of generating a design matrix by collecting multiple sets of the first dimensional data generated for training and the third dimensional data which is a combination of the second dimensional data generated for training, The process of generating fourth-dimensional data from multiple degradation states measured for learning is performed, The model is constructed using the design matrix generated for training and the fourth-dimensional data generated for training. The estimation controller is, A process for generating first-dimensional data having predetermined quantities of feature quantities from the measured value information measured for estimation, The process involves generating second-dimensional data having a predetermined number of feature quantities from the usage history information acquired for estimation, and then executing the following: The battery degradation state estimation method according to claim 1, wherein the degradation state is estimated from a third-dimensional data obtained by combining the first-dimensional data generated for estimation and the second-dimensional data generated for estimation, using the aforementioned model. [Claim 6] The battery degradation state estimation method according to claim 1, wherein the learning controller constructs the model for each type of battery cell. [Claim 7] The battery degradation state estimation method according to claim 1, wherein at least one of the learning controller and the estimation controller is mounted on a vehicle. [Claim 8] The battery degradation state estimation method according to claim 1, wherein at least one of the learning controller and the estimation controller is implemented on an external server capable of communicating with the vehicle. [Claim 9] The battery degradation state estimation method according to claim 1, wherein the estimation controller estimates the degradation state of the battery to be estimated that has been removed from the vehicle. [Claim 10] A process to measure the AC resistance of a battery used for vehicle propulsion by electrochemical impedance spectroscopy, and to obtain measurement information including at least the real axis component of the AC resistance measured for learning, A process to acquire usage history information of the battery to be studied for learning purposes, A process for measuring the degradation state of the battery to be studied for learning purposes, The system includes a learning controller that performs the process of constructing a model in which the measured value information and the usage history information acquired for learning are used as explanatory variables, and the degradation state measured for learning is used as the target variable. A process to measure the AC resistance value of the battery to be estimated for vehicle propulsion using electrochemical impedance spectroscopy, and to obtain measurement information including at least the real axis component of the AC resistance value measured for estimation, A process to acquire usage history information of the battery to be estimated for estimation, The system includes an estimation controller that performs the following steps: using the aforementioned model, the measurement data measured for estimation and the usage history data acquired for estimation are used as explanatory variables, and the degradation state of the battery to be estimated is used as the objective variable. The aforementioned usage history information is characterized by being time-series data in which battery discharge and charging are each treated as separate events, and each event has one or more feature quantities. [Claim 11] The battery degradation state estimation method according to claim 1, characterized in that the usage history information indicates how the battery was used. [Claim 12] The battery degradation state estimation method according to claim 1, characterized in that the usage history information is time-series data indicating how the battery was used.

Images