Information processing method, information processing device, and information processing program

Abstract

This information processing method comprises: acquiring history information including a cumulative elapsed time for each predetermined period of a prediction target battery used in each of a plurality of first use conditions represented by a combination of SOC in a predetermined range and a temperature in a predetermined range and a cumulative charge / discharge amount for each predetermined period of the prediction target battery used in each of a plurality of second use conditions represented by a combination of SOC in a predetermined range and a current rate in a predetermined range; and predicting the deterioration amount for each predetermined period of the prediction target battery by inputting the acquired history information to a learning model.

Description

Information processing method, information processing device, and information processing program

[0001] The present disclosure relates to a technique for predicting the degree of deterioration of a battery.

[0002] For example, Patent Literature 1 discloses a battery degradation diagnosis system that estimates the capacity maintenance rate of a target battery by inputting the usage pattern of the target battery into a capacity estimation model, acquiring the capacity degradation rate of the target battery per unit time, and integrating the acquired capacity degradation rates per unit time. The capacity estimation model is a trained model that uses the usage pattern of the battery regarding the current, SOC, and temperature, and the measured value of the battery's capacity degradation rate per unit time as training data.

[0003] However, with the above-described conventional technology, there is a risk that the accuracy of predicting the deterioration of the target battery may decrease, and further improvement is needed.

[0004] JP 2022-44943 A

[0005] The present disclosure has been made to solve the above problems, and aims to provide a technology that can improve the accuracy of predicting the deterioration of a battery that is a target of prediction.

[0006] An information processing method according to one aspect of the present disclosure is an information processing method executed by a computer, and includes acquiring historical information including the cumulative elapsed time for each predetermined period of a battery to be predicted used under each of a plurality of first usage conditions represented by a combination of a predetermined range of SOC and a predetermined range of temperature, and the cumulative charge / discharge amount for each predetermined period of the battery to be predicted used under each of a plurality of second usage conditions represented by a combination of a predetermined range of SOC and a predetermined range of current rate, and predicting the amount of deterioration of the battery to be predicted for each predetermined period by inputting the acquired historical information into a learning model.

[0007] According to the present disclosure, it is possible to improve the accuracy of predicting the deterioration of a prediction target battery.

[0008] FIG. 1 is a diagram illustrating a configuration of an information processing system according to the present embodiment. FIG. 2 is a diagram illustrating an example of a first feature amount in the present embodiment. FIG. 3 is a flowchart illustrating a prediction process by the information processing device according to the present embodiment. FIG. 4 is a flowchart illustrating a first learning process of a first machine learning model by the information processing device according to the present embodiment. FIG. 5 is a flowchart illustrating a second learning process of a second machine learning model by the information processing device according to the present embodiment. FIG. 6 is a schematic diagram illustrating a first learning process of a first machine learning model in a modified example of the present embodiment.

[0009] (Findings underlying the present disclosure) Battery degradation prediction involves modeling how battery performance deteriorates during use, and taking into account factors such as the battery's chemical reaction, temperature, charge / discharge cycles, and usage conditions.

[0010] Conventional battery degradation prediction is performed using an electrochemical approach based on information on battery materials such as the positive electrode, negative electrode, and electrolyte, but material information is difficult for anyone other than battery manufacturers to obtain. Also, collecting data on battery degradation for each product requires time and testing costs.

[0011] In the above-described conventional technology, the current I, SOC, and temperature T that have been preprocessed are input to a capacity estimation model (neural network model), and the capacity deterioration rate per unit time is output.

[0012] Furthermore, the above-mentioned conventional technology calculates the capacity degradation rate per unit time by dividing the amount of capacity degradation at Δt by Δt. In this case, unless it is assumed that the battery is always used continuously in the same usage pattern, there is a risk that training data cannot be created.

[0013] When creating a capacity estimation model from the usage history of a battery product used by a user, the values ​​of the current I, SOC, and temperature T vary widely, making it difficult to link the usage pattern with the capacity deterioration rate per unit time.

[0014] Furthermore, batteries deteriorate according to a root law, whereby they deteriorate rapidly at first and then gradually deteriorate over time. However, no prior art has been developed that reflects this root law in its deterioration predictions.

[0015] In order to solve the above problems, the following techniques are disclosed.

[0016] (1) An information processing method according to one aspect of the present disclosure is an information processing method executed by a computer, and includes acquiring historical information including the cumulative elapsed time for each predetermined period of a battery to be predicted used under each of a plurality of first usage conditions represented by a combination of a predetermined range of SOC and a predetermined range of temperature, and the cumulative charge / discharge amount for each predetermined period of the battery to be predicted used under each of a plurality of second usage conditions represented by a combination of a predetermined range of SOC and a predetermined range of current rate, and predicting the amount of deterioration of the battery to be predicted for each predetermined period by inputting the acquired historical information into a learning model.

[0017] According to this configuration, historical information including the cumulative elapsed time for each predetermined period of the battery to be predicted used under each of a plurality of first usage conditions represented by a combination of a predetermined range of SOC and a predetermined range of temperature, and the cumulative charge / discharge amount for each predetermined period of the battery to be predicted used under each of a plurality of second usage conditions represented by a combination of a predetermined range of SOC and a predetermined range of current rate, is input into a learning model, and the amount of deterioration for each predetermined period of the battery to be predicted is predicted.

[0018] Therefore, since the deterioration amount is predicted according to various usage conditions of the battery to be predicted, it becomes possible to predict deterioration that reflects the root rule, thereby improving the accuracy of predicting the deterioration of the battery to be predicted.

[0019] (2) The information processing method according to (1) above may further include calculating a predicted deterioration amount at the time of prediction based on the predicted deterioration amount for each of the predetermined periods.

[0020] According to this configuration, the predicted deterioration amount at the time of prediction can be calculated by subtracting the predicted deterioration amount for each predetermined period from the initial time to the time of prediction from the initial deterioration amount.

[0021] (3) In the information processing method described in (1) or (2) above, the learning model may include a first learning model learned based on first historical information of a first battery of the same type as the battery to be predicted, and may further include determining whether the first historical information satisfies a predetermined condition, and predicting the amount of deterioration may include predicting the amount of deterioration of the battery to be predicted for each predetermined period using the first learning model if the first historical information satisfies the predetermined condition.

[0022] According to this configuration, the deterioration amount of the prediction target battery for each predetermined period is predicted using a first learning model learned based on first history information of a first battery that is the same type as the prediction target battery.

[0023] Therefore, although it is difficult to collect a large amount of first history information for a first battery that is the same type as the battery to be predicted, if there is sufficient first history information, the deterioration amount of the battery to be predicted for each specified period can be predicted using a first learning model learned based on the first history information, thereby making it possible to predict the deterioration amount of the battery to be predicted with high accuracy.

[0024] (4) In the information processing method described in (3) above, the learning model may include a second learning model learned based on second historical information of a second battery that is a different type from the battery to be predicted, and predicting the amount of deterioration may include predicting the amount of deterioration of the battery to be predicted for each specified period using the second learning model when the first historical information does not satisfy the specified condition.

[0025] According to this configuration, the deterioration amount of the target battery for prediction for each specified period is predicted using a second learning model learned based on second history information of a second battery that is a different type from the target battery for prediction.

[0026] Therefore, since it is easy to collect a large amount of second history information for a second battery that is a different type from the battery to be predicted, if there is insufficient first history information, the deterioration amount of the battery to be predicted for each specified period is predicted using a second learning model learned based on the second history information, and although the prediction accuracy decreases, the deterioration amount of the battery to be predicted can be reliably predicted.

[0027] (5) In the information processing method described in (3) or (4) above, the predetermined condition may include a condition that the number of records included in the first history information for each predetermined period is equal to or greater than a threshold value.

[0028] According to this configuration, if the number of records for each specified period contained in the first history information is greater than or equal to a threshold value, the deterioration level of the battery to be predicted can be predicted with high accuracy using the first learning model learned based on the first history information.

[0029] (6) In the information processing method described in (3) or (4) above, the predetermined condition may include a condition that the measured amount of deterioration included in the first history information is equal to or less than a predetermined amount.

[0030] According to this configuration, when the measured amount of deterioration included in the first history information is less than a predetermined amount, the amount of deterioration of the battery to be predicted can be predicted with high accuracy using the first learning model learned based on the first history information.

[0031] (7) In the information processing method described above in (4), the cumulative charge / discharge amount included in the second history information of the second battery may be normalized according to a rated capacity of the second battery.

[0032] According to this configuration, the cumulative charge / discharge amount included in the second history information of the second battery is normalized according to the rated capacity of the second battery, so that a highly versatile second learning model can be created, and the prediction accuracy of the deterioration amount of the battery to be predicted can be improved even if the rated capacity of the second battery differs from the rated capacity of the battery to be predicted.

[0033] (8) In the information processing method described in any one of (1) to (7) above, the method may further include acquiring learning history information including the cumulative elapsed time for each predetermined period of the learning battery used under each of the plurality of first usage conditions, the cumulative charge / discharge amount for each predetermined period of the learning battery used under each of the plurality of second usage conditions, and the measured amount of deterioration for each predetermined period; and training the learning model using the cumulative elapsed time for each predetermined period of the learning battery used under each of the plurality of first usage conditions and the cumulative charge / discharge amount for each predetermined period of the learning battery used under each of the plurality of second usage conditions as explanatory variables, and the measured amount of deterioration for each predetermined period of the learning battery as a dependent variable.

[0034] According to this configuration, when the cumulative elapsed time for each predetermined period of a learning battery used under each of a plurality of first usage conditions and the cumulative charge / discharge amount for each predetermined period of a learning battery used under each of a plurality of second usage conditions are input, the learning model can be trained to output the measured deterioration amount of the learning battery for each predetermined period.

[0035] (9) In the information processing method described in (8) above, the predetermined period is a measurement interval for the amount of deterioration, and the method may further include generating corrected history information based on the learning history information by dividing the period from the initial point to the latest point into correction periods that are shorter and equally spaced than the measurement interval; predicting the amount of deterioration for each correction period by inputting the correction history information into the learning model; correcting the predicted amount of deterioration for each correction period at a ratio that matches the amount of deterioration for each predetermined period; and learning the learning model using the cumulative elapsed time for each correction period of the learning battery used under each of the plurality of first usage conditions and the cumulative charge / discharge amount for each correction period of the learning battery used under each of the plurality of second usage conditions as explanatory variables, and the corrected amount of deterioration for each correction period of the learning battery as a dependent variable.

[0036] According to this configuration, the deterioration amount for each correction period, which is equal to or shorter than the measurement interval, is corrected at a ratio that matches the deterioration amount for each predetermined period. A learning model is trained using the cumulative elapsed time for each correction period of the learning battery used under each of a plurality of first usage conditions and the cumulative charge / discharge amount for each correction period of the learning battery used under each of a plurality of second usage conditions as explanatory variables, and the corrected deterioration amount for each correction period of the learning battery as a target variable.

[0037] Therefore, the deterioration amount for each correction period, which is shorter and at equal intervals than the measurement interval, is used for learning the learning model, making it possible to predict the deterioration amount at shorter intervals.

[0038] Furthermore, the present disclosure can be realized not only as an information processing method that executes the characteristic processes described above, but also as an information processing device having a characteristic configuration corresponding to the characteristic processes executed by the information processing method. Furthermore, the present disclosure can also be realized as a computer program that causes a computer to execute the characteristic processes included in such an information processing method. Therefore, the same effects as those of the above information processing method can also be achieved in the following other aspects.

[0039] (10) Another aspect of the present disclosure is an information processing device that includes a processor, and the processor acquires historical information including the cumulative elapsed time for each predetermined period of a battery to be predicted used under each of a plurality of first usage conditions represented by a combination of a predetermined range of SOC and a predetermined range of temperature, and the cumulative charge / discharge amount for each predetermined period of the battery to be predicted used under each of a plurality of second usage conditions represented by a combination of a predetermined range of SOC and a predetermined range of current rate, and predicts the amount of deterioration of the battery to be predicted for each predetermined period by inputting the acquired historical information into a learning model.

[0040] (11) An information processing program according to another aspect of the present disclosure acquires historical information including the cumulative elapsed time for each predetermined period of a battery to be predicted used under each of a plurality of first usage conditions represented by a combination of a predetermined range of SOC and a predetermined range of temperature, and the cumulative charge / discharge amount for each predetermined period of the battery to be predicted used under each of a plurality of second usage conditions represented by a combination of a predetermined range of SOC and a predetermined range of current rate, and causes a computer to function to predict the amount of deterioration of the battery to be predicted for each predetermined period by inputting the acquired historical information into a learning model.

[0041] A non-transitory computer-readable recording medium according to another aspect of the present disclosure records the information processing program described in (11) above.

[0042] Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. Note that each of the embodiments described below represents a specific example of the present disclosure. The numerical values, shapes, components, steps, and order of steps shown in the following embodiments are merely examples and are not intended to limit the present disclosure. Furthermore, among the components in the following embodiments, components that are not described in the independent claims that represent the highest concept are described as optional components. Furthermore, in all embodiments, the respective contents can be combined.

[0043] FIG. 1 is a diagram showing a configuration of an information processing system 1 according to the present embodiment.

[0044] The information processing system 1 includes a prediction target battery 10 , a first battery 11 , a second battery 12 , a third battery 13 , an information processing device 20 , and a terminal 30 .

[0045] The terminal 30 is, for example, a personal computer, a smartphone, or a tablet computer, and displays information to be presented to a user. The terminal 30 is connected to the information processing device 20 via a network so that they can communicate with each other. The network is a local area network or a wide area network. The terminal 30 may be connected to the information processing device 20 via a wired or wireless connection so that they can communicate with each other.

[0046] The prediction target battery 10 is a secondary battery for which degradation is predicted, and stores power by charging and supplies power by discharging. The secondary battery is, for example, a lithium-ion battery. The prediction target battery 10 is installed as a power source in various devices. The prediction target battery 10 is connected to an information processing device 20 via a network so that they can communicate with each other.

[0047] Lithium-ion batteries are classified into several types depending on the material used in the positive electrode. The battery 10 to be predicted may be any of a cobalt-based lithium-ion battery, a manganese-based lithium-ion battery, a nickel-based lithium-ion battery, an NCA-based lithium-ion battery, an iron phosphate-based lithium-ion battery, and a ternary lithium-ion battery.

[0048] The secondary battery is not limited to a lithium ion battery, but may be a lead storage battery, a nickel-metal hydride battery, an all-solid-state battery, or the like.

[0049] The prediction target battery 10 transmits its usage history to the information processing device 20. The usage history includes the date and time, current value, voltage value, capacity, temperature, SOC (State of Charge), and SOH (State of Health). The prediction target battery 10 periodically measures the current value, voltage value, capacity, temperature, SOC, and SOH.

[0050] The current value, voltage value, and capacity of the prediction target battery 10 are measured by a measuring device (not shown) provided in the prediction target battery 10. The temperature of the prediction target battery 10 is measured by a temperature sensor (not shown) provided in the prediction target battery 10. The SOC is an index representing the charging rate of the prediction target battery 10. The SOC of the prediction target battery 10 is expressed as (remaining capacity [Ah] / full charge capacity [Ah])*100.

[0051] The SOH is an index that represents the health of the battery 10 to be predicted. The SOH of the battery 10 to be predicted is expressed as (full charge capacity [Ah] at the time of deterioration (at a specified time) / initial full charge capacity [Ah]) * 100. The SOH represents how much the battery has deteriorated from the time the battery was first used until a specified time. The SOH is an example of the amount of deterioration, and is also called the degree of deterioration.

[0052] Note that the frequency of measuring the current value, voltage value, capacity, temperature, and SOC is different from the frequency of measuring the SOH. The frequency of measuring the current value, voltage value, capacity, temperature, and SOC is higher than the frequency of measuring the SOH. For example, the current value, voltage value, capacity, temperature, and SOC are measured every second, every 10 seconds, or every minute, and the SOH is measured every week, every two weeks, or every month. Therefore, from the first date and time when the SOH is measured to the second date and time when the SOH is next measured, the SOH measured on the first date and time is included in the usage history. Note that the SOH may also be measured irregularly.

[0053] The first battery 11 is the same type of battery as the prediction target battery 10. The first battery 11 is a secondary battery used for training the first machine learning model, storing power by charging and supplying power by discharging. The secondary battery is, for example, a lithium ion battery. The first battery 11 is installed as a power source in various devices. The first battery 11 is connected to the information processing device 20 via a network so that they can communicate with each other. The first battery 11 is a battery of the same model as the prediction target battery 10 and manufactured by the same manufacturer.

[0054] The first battery 11 transmits a first usage history to the information processing device 20. The first usage history includes the date and time, current value, voltage value, capacity, temperature, SOC, and SOH. The first battery 11 periodically measures the current value, voltage value, capacity, temperature, SOC, and SOH. The content of the first usage history is the same as the content of the usage history of the prediction target battery 10.

[0055] The second battery 12 and the third battery 13 are different types of batteries from the prediction target battery 10. The second battery 12 and the third battery 13 are secondary batteries used for training the second machine learning model, storing power by charging and supplying power by discharging. The secondary batteries are, for example, lithium-ion batteries. The second battery 12 and the third battery 13 are installed as power sources in various devices. The second battery 12 and the third battery 13 are connected to the information processing device 20 via a network so that they can communicate with each other.

[0056] The second battery 12 and the third battery 13 are preferably batteries manufactured by a different manufacturer than the prediction target battery 10, or batteries of a different model manufactured by the same manufacturer as the prediction target battery 10. The material used in the positive electrodes of the second battery 12 and the third battery 13 is preferably the same as the material used in the positive electrode of the prediction target battery 10. The rated capacities of the second battery 12 and the third battery 13 may be the same as or different from the rated capacity of the prediction target battery 10. The shape (rectangular, cylindrical, or laminated) of the second battery 12 and the third battery 13 is preferably the same as the shape of the prediction target battery 10.

[0057] The second battery 12 transmits a second usage history to the information processing device 20. The third battery 13 transmits a third usage history to the information processing device 20. The second usage history and the third usage history include date and time, current value, voltage value, capacity, temperature, SOC, and SOH. The second battery 12 and the third battery 13 periodically measure the current value, voltage value, capacity, temperature, SOC, and SOH. The contents of the second usage history and the third usage history are the same as the contents of the usage history of the prediction target battery 10.

[0058] In the present embodiment, information processing system 1 includes one battery (first battery 11) for training the first machine learning model, but the present disclosure is not particularly limited thereto and may include two or more batteries for training the first machine learning model. Also, in the present embodiment, information processing system 1 includes two batteries (second battery 12 and third battery 13) for training the second machine learning model, but the present disclosure is not particularly limited thereto and may include only one battery (second battery 12) for training the second machine learning model, or may include three or more batteries for training the second machine learning model.

[0059] The information processing device 20 includes a first usage history acquisition unit 101, an estimated usage history creation unit 102, a first feature calculation unit 103, a judgment unit 104, a deterioration prediction unit 105, an output unit 106, a second usage history acquisition unit 111, a second feature calculation unit 112, a first learning unit 113, a third usage history acquisition unit 114, a third feature calculation unit 115, a normalization processing unit 116, a second learning unit 117, a feature memory unit 201, a first machine learning model memory unit 211, and a second machine learning model memory unit 212.

[0060] The information processing device 20 includes at least a computer system including, for example, a control program, a processing circuit such as a processor or logic circuit that executes the control program, and a recording device such as an internal memory or an accessible external memory that stores the control program. Note that the information processing device 20 may be realized, for example, by hardware implementation using the processing circuit, or by execution of a software program stored in the memory by the processing circuit or distributed from an external server, or by a combination of these hardware and software implementations.

[0061] The information processing device 20 may be a server or an edge computer. The configuration of the information processing device 20 may be implemented in the terminal 30. The configuration of the information processing device 20 may be distributed among the terminal 30 and the server.

[0062] The information processing device 20 performs a prediction process for predicting the deterioration amount of the prediction target battery 10 using a first machine learning model or a second machine learning model, a first learning process for learning the first machine learning model, and a second learning process for learning the second machine learning model. The first usage history acquisition unit 101, the estimated usage history creation unit 102, the first feature calculation unit 103, the determination unit 104, the deterioration prediction unit 105, the output unit 106, the feature storage unit 201, the first machine learning model storage unit 211, and the second machine learning model storage unit 212 perform the prediction process. The second usage history acquisition unit 111, the second feature calculation unit 112, the first learning unit 113, the feature storage unit 201, and the first machine learning model storage unit 211 perform the first learning process. The third usage history acquisition unit 114, the third feature calculation unit 115, the normalization processing unit 116, the second learning unit 117, and the second machine learning model storage unit 212 perform the second learning process.

[0063] The first usage history acquisition unit 101 acquires a usage history from the initial stage (the time when use begins) of the prediction target battery 10 to a predetermined past measurement time (the most recent measurement time). The predetermined measurement time may be the most recent measurement time or any other measurement time. A communication unit (not shown) receives the usage history transmitted by the prediction target battery 10. The first usage history acquisition unit 101 acquires the usage history received by the communication unit. The usage history is the usage history of the prediction target battery 10 from the initial stage to the most recent measurement time. The terminal 30 may accept input of the usage history by a user, and the communication unit (not shown) may receive the usage history input by the terminal 30. A memory (not shown) may store the usage history, and the first usage history acquisition unit 101 may acquire the usage history from the memory.

[0064] The estimated usage history creation unit 102 creates an estimated usage history of the prediction target battery 10 from a predetermined measurement point (latest measurement point) to a prediction point based on the usage history acquired by the first usage history acquisition unit 101. For example, when predicting the deterioration amount of the prediction target battery 10 one month after the predetermined measurement point, the estimated usage history creation unit 102 copies the usage history for the past one month from the predetermined measurement point and changes the date and time of the copied one month of usage history to the date and time from the predetermined measurement point to one month after. In this way, the estimated usage history creation unit 102 creates an estimated usage history for the period from the predetermined measurement point to one month after. The estimated usage history creation unit 102 adds the created estimated usage history to the usage history acquired by the first usage history acquisition unit 101 to create a usage history from the beginning (the point at which use begins) to the prediction point.

[0065] The terminal 30 may accept an input of the prediction time from the user. The user may input the date and time of the prediction time, or may input the period from the present to the prediction time.

[0066] Furthermore, in this embodiment, a usage history of the prediction target battery 10 exists, and the first usage history acquisition unit 101 acquires the usage history from the beginning (the time when use begins) of the prediction target battery 10 to a predetermined measurement time (the time when the most recent measurement begins), but at the time when use of the prediction target battery 10 begins, no usage history of the prediction target battery 10 exists, and the first usage history acquisition unit 101 cannot acquire the usage history of the prediction target battery 10. Therefore, when no usage history of the prediction target battery 10 exists, the information processing device 20 may further include a battery usage simulator that simulates the usage history from the beginning (present) to the time of prediction based on usage information that indicates how the prediction target battery 10 is used.

[0067] The usage information includes, for example, the hours that the prediction target battery 10 is used per day, the temperature of the prediction target battery 10 per day, and the charge / discharge amount of the prediction target battery 10 per day. The usage information is input by the user. The battery usage simulator may create a usage history from the initial time (present) to the time of prediction based on the usage information indicating how the prediction target battery 10 is used.

[0068] The first feature amount calculation unit 103 calculates feature amounts related to the storage state and charge / discharge state of the prediction target battery 10 based on the usage history from the initial stage (the time when usage started) to a predetermined measurement time (the most recent measurement time) acquired by the first usage history acquisition unit 101 and the estimated usage history from the predetermined measurement time (the most recent measurement time) to the prediction time created by the estimated usage history creation unit 102. The feature amounts are an example of history information.

[0069] The feature quantities include the cumulative elapsed time for each predetermined period of the prediction target battery 10 used under each of a plurality of first usage conditions represented by a combination of a predetermined range of SOC and a predetermined range of temperature, and the cumulative charge / discharge amount for each predetermined period of the prediction target battery 10 used under each of a plurality of second usage conditions represented by a combination of a predetermined range of SOC and a predetermined range of current rate. The predetermined period is the measurement interval for the degradation amount SOH. Note that the predetermined period in the estimated usage history is not an actually measured measurement interval, but a period created based on past measurement intervals.

[0070] The first characteristic calculation unit 103 calculates a characteristic (history information) including the cumulative elapsed time for each predetermined period of the prediction target battery 10 used under each of a plurality of first usage conditions represented by a combination of a predetermined range of SOC and a predetermined range of temperature, and the cumulative charge / discharge amount for each predetermined period of the prediction target battery 10 used under each of a plurality of second usage conditions represented by a combination of a predetermined range of SOC and a predetermined range of current rate.

[0071] For example, the SOC is divided into a range between 0% and 100% in increments of 5%. The current rate is divided into a range between 0 and less than -1, a range between -1 and 1 in increments of 0.1, and a range of 1 or greater. The temperature is divided into a range of 0 degrees or less, a range between 0°C and 60°C in increments of 5°C, and a range of 60 degrees or greater. Note that the above-mentioned predetermined ranges of SOC, current rate, and temperature are merely examples.

[0072] For example, the first characteristic amount calculation unit 103 calculates, based on the usage history, the cumulative elapsed time of the prediction target battery 10 used under first usage conditions represented by a combination of a temperature range of 20-25 degrees and an SOC range of 90-95%. Also, the first characteristic amount calculation unit 103 calculates, based on the usage history, the cumulative charge / discharge amount of the prediction target battery 10 used under second usage conditions represented by a combination of a current rate range of 0.1-0.2 and an SOC range of 50-55%.

[0073] The feature amount storage unit 201 stores a first feature amount (first history information) calculated from a first usage history of a first battery 11 that is the same type as the prediction target battery 10 .

[0074] The determination unit 104 determines whether the first feature (first history information) of the first battery 11 stored in the feature storage unit 201 satisfies a predetermined condition. Here, the predetermined condition includes a condition that the number of records included in the first feature (first history information) for each predetermined period is equal to or greater than a threshold value.

[0075] In addition, the judgment unit 104 may not only determine whether the number of records included in the table of first feature quantities is greater than or equal to a threshold value, but may also determine whether the number of fields of first feature quantities in which a numerical value exists in the same field as the feature quantity of the battery 10 to be predicted is greater than or equal to a threshold value.

[0076] The predetermined condition may include a condition that the measured state of health (SOH) included in the first feature (first history information) is equal to or less than a predetermined amount. For example, if the measured state of health (SOH) included in the first feature is equal to or less than 60%, it can be said that the first machine learning model has been trained using the first feature sufficient to predict the amount of deterioration.

[0077] The first machine learning model memory unit 211 stores a first machine learning model (first learning model) learned based on the first characteristic amount (first history information) of the first battery 11 that is the same type as the prediction target battery 10.

[0078] The first machine learning model is a machine learning model dedicated to the prediction target battery 10 and can predict the deterioration amount of the prediction target battery 10 with high accuracy. However, creating the first machine learning model requires a large amount of long-term usage history of the first battery 11. Therefore, it is not always possible to predict the deterioration amount of the prediction target battery 10 using the first machine learning model. If the amount of data on the first feature amount of the first battery 11 is small, it may not be possible to predict the deterioration amount using the first machine learning model.

[0079] The second machine learning model storage unit 212 stores a second machine learning model (second learning model) learned based on the second feature amount (second history information) of the second battery 12, which is a different type from the prediction target battery 10, and the third feature amount (third history information) of the third battery 13. Note that the second machine learning model storage unit 212 may also store a second machine learning model (second learning model) learned based on the second feature amount (second history information) of the second battery 12, which is a different type from the prediction target battery 10.

[0080] The second machine learning model is a highly versatile machine learning model that can be applied not only to the prediction target battery 10 but also to various other batteries. The prediction accuracy of the deterioration amount of the prediction target battery 10 using the second machine learning model is lower than the prediction accuracy of the deterioration amount of the prediction target battery 10 using the first machine learning model. However, there is a large amount of usage history for the second battery 12 and the third battery 13, and the second machine learning model can be created in advance. Therefore, it is always possible to predict the deterioration amount of the prediction target battery 10 using the second machine learning model.

[0081] The deterioration prediction unit 105 acquires features (history information) including the cumulative elapsed time for each predetermined period of the prediction target battery 10 used under each of a plurality of first usage conditions represented by a combination of a predetermined range of SOC and a predetermined range of temperature, and the cumulative charge / discharge amount for each predetermined period of the prediction target battery 10 used under each of a plurality of second usage conditions represented by a combination of a predetermined range of SOC and a predetermined range of current rate.

[0082] The deterioration prediction unit 105 predicts the deterioration amount ΔSOH for each predetermined period of the prediction target battery 10 by inputting the feature amounts (history information) calculated by the first feature amount calculation unit 103 into a machine learning model (learning model). The machine learning model includes a first machine learning model (first learning model) trained based on the first feature amounts (first history information) of the first battery 11, which is of the same type as the prediction target battery 10, and a second machine learning model (second learning model) trained based on the second feature amounts (second history information) of the second battery 12, which is of a different type from the prediction target battery 10.

[0083] If the determination unit 104 determines that the first feature amount (first history information) satisfies the predetermined condition, the deterioration prediction unit 105 uses the first machine learning model (first learning model) to predict the deterioration amount ΔSOH for each predetermined period of the prediction target battery 10. Furthermore, if the determination unit 104 determines that the first feature amount (first history information) does not satisfy the predetermined condition, the deterioration prediction unit 105 uses the second machine learning model (second learning model) to predict the deterioration amount ΔSOH for each predetermined period of the prediction target battery 10.

[0084] The deterioration prediction unit 105 calculates the deterioration amount ΔSOH of the prediction target battery 10 for each specified period by inputting the feature values ​​of the prediction target battery 10 calculated by the first feature value calculation unit 103 into the first machine learning model or the second machine learning model.

[0085] The deterioration prediction unit 105 calculates the predicted deterioration amount SOH at the time of prediction based on the predicted deterioration amount ΔSOH for each predetermined period. Ｘ The deterioration prediction unit 105 calculates the initial deterioration amount SOH ０ The predicted deterioration amount SOH at the time of prediction is calculated by subtracting the deterioration amount ΔSOH predicted for each predetermined period from the initial period to the time of prediction from the above. Ｘ Calculate.

[0086] The determination unit 104 may determine whether or not the first machine learning model is stored in the first machine learning model storage unit 211. If the first machine learning model is stored in the first machine learning model storage unit 211, the deterioration prediction unit 105 may acquire the first machine learning model from the first machine learning model storage unit 211. If the first machine learning model is not stored in the first machine learning model storage unit 211, the deterioration prediction unit 105 may acquire the second machine learning model from the second machine learning model storage unit 212.

[0087] The output unit 106 outputs the predicted deterioration amount SOH at the time of prediction calculated by the deterioration prediction unit 105. Ｘ The output unit 106 outputs the predicted deterioration amount SOH via a communication unit (not shown). Ｘ The terminal 30 transmits the predicted deterioration amount SOH of the prediction target battery 10 transmitted by the information processing device 20 to the terminal 30. Ｘ and receives the predicted deterioration amount SOH of the prediction target battery 10. Ｘ This allows the user to know the degree of deterioration of the prediction target battery 10 at the time of prediction, and the user can know the degree of deterioration of the prediction target battery 10 at the time of prediction.

[0088] The second usage history acquisition unit 111 acquires a first usage history of the first battery 11, which is of the same type as the prediction target battery 10, from the initial stage (the time when use begins) to a predetermined measurement time point in the past (the most recent measurement time point). A communication unit (not shown) receives the first usage history transmitted by the first battery 11. The second usage history acquisition unit 111 acquires the first usage history received by the communication unit. The first usage history is the usage history of the first battery 11 from the initial stage to a predetermined measurement time point in the past. Note that the terminal 30 may accept input of the first usage history by a user, and the communication unit (not shown) may receive the first usage history input by the terminal 30. Note that a memory (not shown) may store the first usage history, and the second usage history acquisition unit 111 may acquire the first usage history from the memory.

[0089] The second feature amount calculation unit 112 calculates first feature amounts related to the storage state and charge / discharge state of the first battery 11 based on the first usage history from the initial time (the time when usage started) to a predetermined measurement time point in the past (the most recent measurement time point) acquired by the second usage history acquisition unit 111. The first feature amounts are an example of first history information.

[0090] The first characteristic amount includes a cumulative elapsed time for each predetermined period of time of the first battery 11 used under each of a plurality of first usage conditions expressed by a combination of a predetermined range of SOC and a predetermined range of temperature, and a cumulative charge / discharge amount for each predetermined period of time of the first battery 11 used under each of a plurality of second usage conditions expressed by a combination of a predetermined range of SOC and a predetermined range of current rate. The predetermined period is the measurement interval of the degradation amount SOH.

[0091] The second characteristic calculation unit 112 calculates a first characteristic (first history information) including the cumulative elapsed time for each predetermined period of the first battery 11 used under each of a plurality of first usage conditions represented by a combination of a predetermined range of SOC and a predetermined range of temperature, and the cumulative charge / discharge amount for each predetermined period of the first battery 11 used under each of a plurality of second usage conditions represented by a combination of a predetermined range of SOC and a predetermined range of current rate.

[0092] FIG. 2 is a diagram showing an example of the first feature amount in this embodiment.

[0093] The first characteristic amount is calculated for each predetermined period from the initial time to the time when the SOH is measured for the Nth time. For example, the first characteristic amount is calculated for each period during which the measured SOH in the first usage history changes. In FIG. 2 , the first characteristic amount is calculated for the period from March 10, 2023 to March 21, 2023 and the period from March 22, 2023 to March 29, 2023.

[0094] In the example of Figure 2, during the period from 2023 / 3 / 10 to 2023 / 3 / 21, the cumulative elapsed time at a temperature of 20-25°C and an SOC of 90-95% is 13.4 hours, the cumulative charge / discharge capacity at a current rate of 0.1-0.2C and an SOC of 50-55% is 21 Ah, and the measured SOH is 98.6. Also, during the period from 2023 / 3 / 22 to 2023 / 3 / 29, the cumulative elapsed time at a temperature of 20-25°C and an SOC of 90-95% is 2.4 hours, the cumulative charge / discharge capacity at a current rate of 0.1-0.2C and an SOC of 50-55% is 30 Ah, and the measured SOH is 98.0.

[0095] Furthermore, the second characteristic amount calculation unit 112 calculates the deterioration amount ΔSOH of the first battery 11 for each predetermined period based on the first usage history. For example, if the deterioration amount SOH measured on 2023 / 3 / 10 is 98.6% and the deterioration amount SOH measured on 2023 / 3 / 22 is 98.0%, the deterioration amount ΔSOH for the period from 2023 / 3 / 10 to 2023 / 3 / 21 is 0.6%. That is, the second characteristic amount calculation unit 112 calculates the deterioration amount SOH measured in the Nth period. Ｎ From the (N+1)th period, the deterioration amount SOH measured Ｎ＋１ By subtracting the deterioration amount ΔSOH of the first battery 11 for the Nth period, Ｎ Calculate.

[0096] Furthermore, the second feature amount calculation unit 112 stores the calculated first feature amount (first history information) in the feature amount storage unit 201 .

[0097] The first learning unit 113 acquires first characteristics (learning history information) including the cumulative elapsed time for each predetermined period of the first battery 11 (learning battery) used under each of a plurality of first usage conditions represented by a combination of a predetermined range of SOC and a predetermined range of temperature, the cumulative charge / discharge amount for each predetermined period of the first battery 11 (learning battery) used under each of a plurality of second usage conditions represented by a combination of a predetermined range of SOC and a predetermined range of current rate, and the measured deterioration amount ΔSOH for each predetermined period.

[0098] The first learning unit 113 uses, as explanatory variables, the cumulative elapsed time for each predetermined period of the first battery 11 (learning battery) used under each of a plurality of first usage conditions represented by combinations of a predetermined range of SOC and a predetermined range of temperature, and the cumulative charge / discharge amount for each predetermined period of the first battery 11 (learning battery) used under each of a plurality of second usage conditions represented by combinations of a predetermined range of SOC and a predetermined range of current rate, and uses, as a target variable, the degradation amount ΔSOH measured for each predetermined period of the first battery 11 (learning battery) to learn a first machine learning model (learning model) stored in the first machine learning model storage unit 211. The first learning unit 113 performs learning using, for example, a Light Gradient Boosting Machine (LightGBM).

[0099] As shown in FIG. 2, the explanatory variable is the first feature amount calculated from the first usage history, and the dependent variable is the deterioration amount ΔSOH for each predetermined period.

[0100] The first learning unit 113 stores the learned first machine learning model in the first machine learning model storage unit 211.

[0101] The third usage history acquisition unit 114 acquires a second usage history of the second battery 12, which is a different type from the prediction target battery 10, from the initial stage (the time when use begins) to a predetermined measurement time point in the past (the most recent measurement time point). A communication unit (not shown) receives the second usage history transmitted by the second battery 12. The third usage history acquisition unit 114 acquires the second usage history received by the communication unit. The second usage history is the usage history of the second battery 12 from the initial stage to a predetermined measurement time point in the past. Note that the terminal 30 may accept input of the second usage history by a user, and the communication unit (not shown) may receive the second usage history input by the terminal 30. Note that a memory (not shown) may store the second usage history, and the third usage history acquisition unit 114 may acquire the second usage history from the memory.

[0102] Furthermore, the third usage history acquisition unit 114 acquires a third usage history of the third battery 13, which is a different type from the prediction target battery 10 and the second battery 12, from the initial stage (the time when use begins) to a predetermined measurement time point in the past (the most recent measurement time point). A communication unit (not shown) receives the third usage history transmitted by the third battery 13. The third usage history acquisition unit 114 acquires the third usage history received by the communication unit. The third usage history is the usage history of the third battery 13 from the initial stage to a predetermined measurement time point in the past. Note that the terminal 30 may accept input of the third usage history by a user, and the communication unit (not shown) may receive the third usage history input by the terminal 30. A memory (not shown) may store the third usage history, and the third usage history acquisition unit 114 may acquire the third usage history from the memory.

[0103] The third feature amount calculation unit 115 calculates second feature amounts relating to the storage state and charge / discharge state of the second battery 12 based on the second usage history from the initial time (the time when usage started) to a predetermined past measurement time point acquired by the third usage history acquisition unit 114. The second feature amounts are an example of second history information.

[0104] The second characteristic amount includes a cumulative elapsed time for each predetermined period of time of the second battery 12 used under each of a plurality of first usage conditions expressed by a combination of a predetermined range of SOC and a predetermined range of temperature, and a cumulative charge / discharge amount for each predetermined period of time of the second battery 12 used under each of a plurality of second usage conditions expressed by a combination of a predetermined range of SOC and a predetermined range of current rate. The predetermined period is the measurement interval of the degradation amount SOH.

[0105] The third characteristic calculation unit 115 calculates a second characteristic (second history information) including the cumulative elapsed time for each predetermined period of the second battery 12 used under each of a plurality of first usage conditions represented by a combination of a predetermined range of SOC and a predetermined range of temperature, and the cumulative charge / discharge amount for each predetermined period of the second battery 12 used under each of a plurality of second usage conditions represented by a combination of a predetermined range of SOC and a predetermined range of current rate.

[0106] The second characteristic amount is calculated for each predetermined period from the initial time to the time when the SOH is measured for the Nth time. For example, the second characteristic amount is calculated for each period during which the measured degradation amount SOH changes in the second usage history.

[0107] The third characteristic amount calculation unit 115 also calculates the deterioration amount ΔSOH of the second battery 12 for each predetermined period based on the second usage history.

[0108] The third feature amount calculation unit 115 calculates third feature amounts related to the storage state and charge / discharge state of the third battery 13 based on the third usage history from the initial time (the time when usage started) to a predetermined measurement time point in the past (the most recent measurement time point) acquired by the third usage history acquisition unit 114. The third feature amount is an example of third history information.

[0109] The third characteristic amount includes the cumulative elapsed time for each predetermined period of time of the third battery 13 used under each of a plurality of first usage conditions expressed by a combination of a predetermined range of SOC and a predetermined range of temperature, and the cumulative charge / discharge amount for each predetermined period of time of the third battery 13 used under each of a plurality of second usage conditions expressed by a combination of a predetermined range of SOC and a predetermined range of current rate. The predetermined period is the measurement interval of the degradation amount SOH.

[0110] The third characteristic calculation unit 115 calculates a third characteristic (third history information) including the cumulative elapsed time for each predetermined period of the third battery 13 used under each of a plurality of first usage conditions represented by a combination of a predetermined range of SOC and a predetermined range of temperature, and the cumulative charge / discharge amount for each predetermined period of the third battery 13 used under each of a plurality of second usage conditions represented by a combination of a predetermined range of SOC and a predetermined range of current rate.

[0111] The third characteristic amount is calculated for each predetermined period from the initial time to the time when the SOH is measured for the Nth time. For example, the third characteristic amount is calculated for each period during which the measured degradation amount SOH changes in the third usage history.

[0112] The third characteristic amount calculation unit 115 also calculates the deterioration amount ΔSOH of the third battery 13 for each predetermined period based on the third usage history.

[0113] The normalization processing unit 116 normalizes the cumulative charge / discharge amount included in the second characteristic amount (second history information) of the second battery 12 according to the rated capacity of the second battery 12. The normalization processing unit 116 also normalizes the cumulative charge / discharge amount included in the third characteristic amount (third history information) of the third battery 13 according to the rated capacity of the third battery 13. The rated capacity of the second battery 12 may be acquired from the second battery 12 together with the second usage history of the second battery 12, and the rated capacity of the third battery 13 may be acquired from the third battery 13 together with the third usage history of the third battery 13.

[0114] For example, if the rated capacity of the second battery 12 is 5 Ah and the rated capacity of the third battery 13 is 100 Ah, the amount of electricity flowing in one cycle through the third battery 13 is 20 times the amount of electricity flowing in one cycle through the second battery 12. Therefore, the normalization processing unit 116 divides the cumulative charge / discharge amount included in the second characteristic amount (second history information) of the second battery 12 by the rated capacity of the second battery 12. This allows the units of cumulative charge / discharge amount to be consistent between batteries with different rated capacities.

[0115] The second learning unit 117 acquires second characteristics (learning history information) including the cumulative elapsed time for each predetermined period of the second battery 12 (learning battery) used under each of a plurality of first usage conditions represented by a combination of a predetermined range of SOC and a predetermined range of temperature, the cumulative charge / discharge amount for each predetermined period of the second battery 12 (learning battery) used under each of a plurality of second usage conditions represented by a combination of a predetermined range of SOC and a predetermined range of current rate, and the measured deterioration amount ΔSOH for each predetermined period.

[0116] The second learning unit 117 uses, as explanatory variables, the cumulative elapsed time for each predetermined period of the second battery 12 (learning battery) used under each of a plurality of first usage conditions represented by combinations of a predetermined range of SOC and a predetermined range of temperature, and the cumulative charge / discharge amount for each predetermined period of the second battery 12 (learning battery) used under each of a plurality of second usage conditions represented by combinations of a predetermined range of SOC and a predetermined range of current rate, and uses, as a target variable, the degradation amount ΔSOH measured for each predetermined period of the second battery 12 (learning battery). The second learning unit 117 performs learning using, for example, LightGBM.

[0117] In addition, the second learning unit 117 acquires a third characteristic (learning history information) including the cumulative elapsed time for each predetermined period of the third battery 13 (learning battery) used under each of a plurality of first usage conditions represented by a combination of a predetermined range of SOC and a predetermined range of temperature, the cumulative charge / discharge amount for each predetermined period of the third battery 13 (learning battery) used under each of a plurality of second usage conditions represented by a combination of a predetermined range of SOC and a predetermined range of current rate, and the measured deterioration amount ΔSOH for each predetermined period.

[0118] The second learning unit 117 learns the second machine learning model (learning model) stored in the second machine learning model memory unit 212 using, as explanatory variables, the cumulative elapsed time for each predetermined period of the third battery 13 (learning battery) used under each of a plurality of first usage conditions represented by a combination of a predetermined range of SOC and a predetermined range of temperature, and the cumulative charge / discharge amount for each predetermined period of the third battery 13 (learning battery) used under each of a plurality of second usage conditions represented by a combination of a predetermined range of SOC and a predetermined range of current rate, and the measured deterioration amount ΔSOH of the third battery 13 (learning battery) for each predetermined period as the objective variable.

[0119] The second learning unit 117 stores the learned second machine learning model in the second machine learning model storage unit 212.

[0120] Next, the prediction process performed by the information processing device 20 in this embodiment will be described.

[0121] FIG. 3 is a flowchart for explaining the prediction process performed by the information processing device 20 according to this embodiment.

[0122] First, in step S1, the first usage history acquisition unit 101 acquires the usage history of the prediction target battery 10 from the beginning (the time when usage started) to a predetermined measurement time point in the past (the most recent measurement time point).

[0123] Next, in step S2, the estimated usage history creation unit 102 creates an estimated usage history of the prediction target battery 10 from a specified past measurement point (latest measurement point) to the prediction point based on the usage history acquired by the first usage history acquisition unit 101.

[0124] Next, in step S3, the first characteristic amount calculation unit 103 calculates characteristic amounts of the prediction target battery 10, including the cumulative elapsed time for each predetermined period of the prediction target battery 10 used under each of a plurality of first usage conditions represented by a combination of a predetermined range of SOC and a predetermined range of temperature, and the cumulative charge / discharge amount for each predetermined period of the prediction target battery 10 used under each of a plurality of second usage conditions represented by a combination of a predetermined range of SOC and a predetermined range of current rate, based on the usage history from the initial period (the time when use began) to a predetermined measurement time point in the past (the most recent measurement time point) acquired by the first usage history acquisition unit 101 and the estimated usage history from a predetermined measurement time point in the past to the prediction time point created by the estimated usage history creation unit 102.

[0125] Next, in step S4 , the determination unit 104 acquires the first characteristic amount of the first battery 11 from the characteristic amount storage unit 201 .

[0126] Next, in step S5, the determination unit 104 determines whether the first characteristic amount of the first battery 11 satisfies a predetermined condition. Here, if the number of records per predetermined period included in the first characteristic amount is equal to or greater than a threshold, the determination unit 104 determines that the first characteristic amount of the first battery 11 satisfies the predetermined condition. Furthermore, if the number of records per predetermined period included in the first characteristic amount is less than the threshold, the determination unit 104 determines that the first characteristic amount of the first battery 11 does not satisfy the predetermined condition.

[0127] Here, if it is determined that the first characteristic value of the first battery 11 satisfies the predetermined condition (YES in step S5), in step S6, the deterioration prediction unit 105 obtains from the first machine learning model memory unit 211 a first machine learning model learned based on the first characteristic value of the first battery 11, which is of the same type as the battery 10 to be predicted.

[0128] Next, in step S7, the deterioration prediction unit 105 uses the first machine learning model to calculate the deterioration amount ΔSOH for each predetermined period of the prediction target battery 10. The deterioration prediction unit 105 calculates the deterioration amount ΔSOH for each predetermined period of the prediction target battery 10 by inputting the feature quantities of the prediction target battery 10 calculated by the first feature quantity calculation unit 103 into the first machine learning model.

[0129] On the other hand, if it is determined that the first characteristic value of the first battery 11 does not satisfy the specified condition (NO in step S5), in step S8, the deterioration prediction unit 105 obtains from the second machine learning model memory unit 212 a second machine learning model learned based on the second characteristic value of the second battery 12, which is of a different type from the battery 10 to be predicted.

[0130] Next, in step S9, the deterioration prediction unit 105 uses the second machine learning model to calculate the deterioration amount ΔSOH for each predetermined period of the prediction target battery 10. The deterioration prediction unit 105 calculates the deterioration amount ΔSOH for each predetermined period of the prediction target battery 10 by inputting the feature quantities of the prediction target battery 10 calculated by the first feature quantity calculation unit 103 into the second machine learning model.

[0131] Next, in step S10, the deterioration prediction unit 105 calculates the initial deterioration amount SOH ０ The predicted deterioration amount SOH at the time of prediction is calculated by subtracting the deterioration amount ΔSOH predicted for each predetermined period from the initial period to the time of prediction from the above. Ｘ Calculate.

[0132] Next, in step S11, the output unit 106 outputs the predicted deterioration amount SOH at the time of prediction calculated by the deterioration prediction unit 105. Ｘ Output.

[0133] In this way, historical information including the cumulative elapsed time for each specified period of the battery 10 to be predicted used under each of a plurality of first usage conditions represented by a combination of a predetermined range of SOC and a predetermined range of temperature, and the cumulative charge / discharge amount for each specified period of the battery 10 to be predicted used under each of a plurality of second usage conditions represented by a combination of a predetermined range of SOC and a predetermined range of current rate, is input into the learning model, and the amount of deterioration for each specified period of the battery 10 to be predicted is predicted.

[0134] Therefore, since the amount of deterioration can be predicted according to various usage conditions of the battery 10 to be predicted, it becomes possible to predict deterioration that reflects the root rule, thereby improving the accuracy of predicting the deterioration of the battery 10 to be predicted.

[0135] Next, the first learning process of the first machine learning model performed by the information processing device 20 in this embodiment will be described.

[0136] FIG. 4 is a flowchart for explaining the first learning process of the first machine learning model by the information processing device 20 in this embodiment.

[0137] First, in step S21, the second usage history acquisition unit 111 acquires the first usage history of the first battery 11, which is of the same type as the prediction target battery 10, from the initial stage (the point at which use begins) to a predetermined measurement point in the past (the most recent measurement point).

[0138] Next, in step S22, the second characteristic amount calculation unit 112 calculates first characteristic amounts of the first battery 11, including the cumulative elapsed time for each predetermined period of the first battery 11 used under each of a plurality of first usage conditions represented by a combination of a predetermined range of SOC and a predetermined range of temperature, and the cumulative charge / discharge amount for each predetermined period of the first battery 11 used under each of a plurality of second usage conditions represented by a combination of a predetermined range of SOC and a predetermined range of current rate, based on the first usage history acquired by the second usage history acquisition unit 111 from the initial period (the time when use began) to a predetermined measurement time point in the past (the most recent measurement time point).

[0139] Next, in step S23 , the second characteristic amount calculation unit 112 stores the calculated first characteristic amount of the first battery 11 in the characteristic amount storage unit 201 .

[0140] Next, in step S24, the second characteristic amount calculation unit 112 calculates the deterioration amount ΔSOH of the first battery 11 for each predetermined period based on the first usage history.

[0141] Next, in step S25, the first learning unit 113 learns the first machine learning model stored in the first machine learning model memory unit 211 using the first characteristic quantity of the first battery 11 as an explanatory variable and the deterioration amount ΔSOH of the first battery 11 per specified period as a target variable.

[0142] Next, in step S26, the first learning unit 113 stores the learned first machine learning model in the first machine learning model storage unit 211.

[0143] Next, the second learning process of the second machine learning model performed by the information processing device 20 in this embodiment will be described.

[0144] FIG. 5 is a flowchart for explaining the second learning process of the second machine learning model by the information processing device 20 in this embodiment.

[0145] First, in step S31, the third usage history acquisition unit 114 acquires the second usage history of the second battery 12, which is of a different type from the prediction target battery 10, from the initial stage (the point at which use begins) to a predetermined measurement point in the past (the most recent measurement point).

[0146] Next, in step S32, the third usage history acquisition unit 114 acquires the third usage history of the third battery 13, which is a different type from the prediction target battery 10 and the second battery 12, from the initial stage (the point at which use begins) to a specified measurement point in the past (the most recent measurement point).

[0147] Next, in step S33, the third characteristic amount calculation unit 115 calculates second characteristic amounts of the second battery 12, including the cumulative elapsed time for each predetermined period of the second battery 12 used under each of a plurality of first usage conditions represented by a combination of a predetermined range of SOC and a predetermined range of temperature, and the cumulative charge / discharge amount for each predetermined period of the second battery 12 used under each of a plurality of second usage conditions represented by a combination of a predetermined range of SOC and a predetermined range of current rate, based on the second usage history acquired by the third usage history acquisition unit 114 from the initial period (the time when use began) to a predetermined measurement time point in the past (the most recent measurement time point).

[0148] Next, in step S34, the third characteristic amount calculation unit 115 calculates third characteristic amounts of the third battery 13, including the cumulative elapsed time for each predetermined period of the third battery 13 used under each of a plurality of first usage conditions represented by a combination of a predetermined range of SOC and a predetermined range of temperature, and the cumulative charge / discharge amount for each predetermined period of the third battery 13 used under each of a plurality of second usage conditions represented by a combination of a predetermined range of SOC and a predetermined range of current rate, based on the third usage history acquired by the third usage history acquisition unit 114 from the initial period (the time when use began) to a predetermined measurement time point in the past (the most recent measurement time point).

[0149] Next, in step S35, the third characteristic amount calculation unit 115 calculates the deterioration amount ΔSOH of the second battery 12 for each predetermined period based on the second usage history.

[0150] Next, in step S36, the third characteristic amount calculation unit 115 calculates the deterioration amount ΔSOH of the third battery 13 for each predetermined period based on the third usage history.

[0151] Next, in step S37 , the normalization processing unit 116 normalizes the cumulative charge / discharge amount included in the second characteristic amount of the second battery 12 according to the rated capacity of the second battery 12 .

[0152] Next, in step S38 , the normalization processing unit 116 normalizes the cumulative charge / discharge amount included in the third characteristic amount of the third battery 13 according to the rated capacity of the third battery 13 .

[0153] Next, in step S39, the second learning unit 117 learns the second machine learning model stored in the second machine learning model memory unit 212 using the second characteristic quantity of the second battery 12 as an explanatory variable and the deterioration amount ΔSOH of the second battery 12 per specified period as a target variable.

[0154] Next, in step S40, the second learning unit 117 learns the second machine learning model stored in the second machine learning model memory unit 212 using the third feature value of the third battery 13 as an explanatory variable and the deterioration amount ΔSOH of the third battery 13 per specified period as a target variable.

[0155] Next, in step S41, the second learning unit 117 stores the learned second machine learning model in the second machine learning model storage unit 212.

[0156] Next, a modification of the present embodiment will be described. In the present embodiment, the first learning unit 113 uses, as explanatory variables, first feature quantities of the first battery 11, including the cumulative elapsed time per predetermined period of the first battery 11 used under each of a plurality of first usage conditions represented by a combination of a predetermined range of SOC and a predetermined range of temperature, and the cumulative charge / discharge amount per predetermined period of the first battery 11 used under each of a plurality of second usage conditions represented by a combination of a predetermined range of SOC and a predetermined range of current rate, and trains a first machine learning model using the degradation amount ΔSOH of the first battery 11 per predetermined period as a target variable. In this case, the predetermined period is the measurement interval for measuring the degradation amount SOH. In contrast, in a modification of the present embodiment, first corrected feature quantities are generated by dividing the period from the initial point to the latest point into correction periods that are shorter than the measurement interval and at equal intervals, based on the first feature quantities.

[0157] FIG. 6 is a schematic diagram for explaining the first learning process of the first machine learning model in a modified example of this embodiment.

[0158] In a modification of the present embodiment, the information processing device 20 may further include a first feature amount corrector, a second deterioration predictor, and a prediction result corrector.

[0159] The first learning unit 113 learns a first machine learning model 303 using a first characteristic quantity 301 of the first battery 11 as an explanatory variable, which includes the cumulative elapsed time for each predetermined period of the first battery 11 used under each of a plurality of first usage conditions represented by a combination of a predetermined range of SOC and a predetermined range of temperature, and the cumulative charge / discharge amount for each predetermined period of the first battery 11 used under each of a plurality of second usage conditions represented by a combination of a predetermined range of SOC and a predetermined range of current rate, and using the deterioration amount ΔSOH302 of the first battery 11 for each predetermined period as a target variable.

[0160] 6, for a first period from 2023 / 3 / 10 to 2023 / 3 / 21, the cumulative elapsed time at a temperature of 20-25°C and an SOC of 90-95%, the cumulative charge / discharge amount at a current rate of 0.1-0.2C and an SOC of 50-55%, and the degradation amount ΔSOH for the first period are calculated. Also, for a second period from 2023 / 3 / 22 to 2023 / 3 / 29, the cumulative elapsed time at a temperature of 20-25°C and an SOC of 90-95%, the cumulative charge / discharge amount at a current rate of 0.1-0.2C and an SOC of 50-55%, and the degradation amount ΔSOH for the second period are calculated.

[0161] The first feature amount correcting unit generates first corrected feature amounts (correction history information) 304 by dividing the period from the initial point to the latest point into correction periods that are equal intervals and shorter than the measurement interval, based on the first feature amounts (learning history information) 301 calculated by the second feature amount calculating unit 112. The correction period is, for example, one day.

[0162] In the example of FIG. 6, the cumulative elapsed time at a temperature of 20-25° C. and an SOC of 90-95% and the cumulative charge / discharge amount at a current rate of 0.1-0.2 C and an SOC of 50-55% are calculated for each day.

[0163] The second deterioration prediction unit predicts the deterioration amount ΔSOH305 for each correction period by inputting the first corrected feature (correction history information) 304 generated by the first feature correction unit into the first machine learning model (learning model) 303.

[0164] 6 , a daily deterioration amount ΔSOH 305 is predicted. Since the first machine learning model 303 is trained using the first feature amount 301 for each predetermined period, the accuracy of the deterioration amount ΔSOH 305 predicted using the first machine learning model 303 is low.

[0165] The prediction result corrector corrects the deterioration amount ΔSOH305 for each correction period predicted by the second deterioration predictor at a ratio that matches the deterioration amount ΔSOH302 for each predetermined period.

[0166] 6 , the measured value of the degradation amount ΔSOH302 for the period from 2023 / 3 / 10 to 2023 / 3 / 21 is 0.6, and the predicted values ​​of the degradation amount ΔSOH305 for each day from 2023 / 3 / 10 to 2023 / 3 / 21 are 0.020, 0.014, ..., 0.009. The prediction result correction unit calculates the correction ratio by dividing the measured value of the degradation amount ΔSOH302 for the period from 2023 / 3 / 10 to 2023 / 3 / 21 by the sum of the predicted values ​​of the degradation amount ΔSOH305 for each day from 2023 / 3 / 10 to 2023 / 3 / 21. The prediction result correction unit then corrects the predicted value of the daily deterioration amount ΔSOH305 by multiplying each predicted value of the daily deterioration amount ΔSOH305 from March 10, 2023 to March 21, 2023 by the correction ratio. This calculates a corrected predicted value of the daily deterioration amount ΔSOH306. Since the predicted daily deterioration amount ΔSOH305 is corrected using the measured deterioration amount ΔSOH302 for the predetermined period, the accuracy of the corrected daily deterioration amount ΔSOH306 is high.

[0167] The first learning unit 113 then learns a first machine learning model (learning model) 303 using, as an explanatory variable, a first correction feature 304 of the first battery 11, which includes the cumulative elapsed time for each correction period of the first battery 11 (learning battery) used under each of a plurality of first usage conditions and the cumulative charge / discharge amount for each correction period of the first battery 11 (learning battery) used under each of a plurality of second usage conditions, and using, as a target variable, the corrected deterioration amount ΔSOH306 of the first battery 11 (learning battery) for each correction period.

[0168] In a modification of this embodiment, a second machine learning model may be trained in the same manner as the first machine learning model. That is, based on the second feature calculated by the third feature calculation unit 115, second corrected feature quantities may be generated by dividing the period from the initial point to the latest point into correction periods that are equal intervals and shorter than the measurement interval. The generated second corrected feature quantities may be input to the second machine learning model to predict the deterioration amount ΔSOH for each correction period. The predicted deterioration amount ΔSOH for each correction period may then be corrected at a ratio that matches the deterioration amount ΔSOH for each predetermined period. The second learning unit 117 may train the second machine learning model using, as explanatory variables, the second corrected feature quantities of the second battery 12, including the cumulative elapsed time for each correction period of the second battery 12 used under each of the plurality of first usage conditions and the cumulative charge / discharge amount for each correction period of the second battery 12 used under each of the plurality of second usage conditions, and the corrected deterioration amount ΔSOH for each correction period of the second battery 12 as a target variable.

[0169] The measurement interval of the degradation amount used as training data for the machine learning model depends on the historical data. Because battery usage methods differ from user to user, there are many variations to be learned. In other words, a large amount of training data is required. In contrast, in a modified example of this embodiment, the measurement interval of the degradation amount is standardized to, for example, one day, thereby reducing the amount of training data required. Furthermore, it becomes possible to predict the degradation amount at shorter intervals.

[0170] Note that some or all of the functions of the device according to the embodiment of the present disclosure may be realized by a processor such as a CPU executing a program.

[0171] Furthermore, all the numbers used above are merely examples to specifically explain the present disclosure, and the present disclosure is not limited to the numbers used as examples.

[0172] The order in which the steps are executed in the above flowchart is merely an example for specifically explaining the present disclosure, and other orders may be used as long as similar effects are obtained. Also, some of the steps may be executed simultaneously (in parallel) with other steps.

[0173] The technology according to the present disclosure can improve the accuracy of predicting the deterioration of a prediction target battery, and is therefore useful as a technology for predicting the amount of deterioration of a battery.

Claims

1. A computer-implemented information processing method, comprising: acquiring history information including an accumulated elapsed time for each predetermined period of a prediction target battery used under each of a plurality of first usage conditions expressed by a combination of a predetermined range of SOC and a predetermined range of temperature, and an accumulated charge / discharge amount for each predetermined period of the prediction target battery used under each of a plurality of second usage conditions expressed by a combination of a predetermined range of SOC and a predetermined range of current rate; predicting the deterioration amount of the prediction target battery for each predetermined period by inputting the acquired history information into a learning model; An information processing method including: 　 and further comprising calculating a predicted deterioration amount at a time point of prediction based on the predicted deterioration amount for each predetermined period. The information processing method according to claim 1. 　 the learning model includes a first learning model learned based on first history information of a first battery that is the same type as the prediction target battery; The method further includes determining whether the first history information satisfies a predetermined condition; predicting the deterioration amount includes predicting the deterioration amount of the prediction target battery for each predetermined period using the first learning model when the first history information satisfies the predetermined condition. The information processing method according to claim 1. 　 the learning model includes a second learning model learned based on second history information of a second battery that is a different type from the prediction target battery; predicting the deterioration amount includes predicting the deterioration amount of the prediction target battery for each predetermined period using the second learning model when the first history information does not satisfy the predetermined condition.

4. The information processing method according to claim 3. 　 the predetermined condition includes a condition that the number of records included in the first history information for each predetermined period is equal to or greater than a threshold value; 5. The information processing method according to claim 3 or 4. 　 the predetermined condition includes a condition that the measured deterioration amount included in the first history information is equal to or less than a predetermined amount.

5. The information processing method according to claim 3 or 4. 　 the cumulative charge / discharge amount included in the second history information of the second battery is normalized according to a rated capacity of the second battery; 5. The information processing method according to claim 4. 　 acquiring learning history information including the cumulative elapsed time for each predetermined period of the learning battery used under each of the plurality of first usage conditions, the cumulative charge / discharge amount for each predetermined period of the learning battery used under each of the plurality of second usage conditions, and the measured amount of deterioration for each predetermined period; training the learning model using, as explanatory variables, the cumulative elapsed time for each predetermined period of the learning battery used under each of the plurality of first usage conditions and the cumulative charge / discharge amount for each predetermined period of the learning battery used under each of the plurality of second usage conditions, and the measured deterioration amount for each predetermined period of the learning battery as a response variable; further comprising: The information processing method according to any one of claims 1 to 4. 　 the predetermined period is a measurement interval of the deterioration amount, generating correction history information by dividing a period from an initial point to a latest point into correction periods that are shorter than the measurement interval and at equal intervals based on the learning history information; predicting the deterioration amount for each correction period by inputting the correction history information into the learning model; correcting the predicted deterioration amount for each correction period at a ratio that matches the deterioration amount for each predetermined period; training the learning model using, as explanatory variables, the cumulative elapsed time for each correction period of the learning battery used under each of the plurality of first usage conditions and the cumulative charge / discharge amount for each correction period of the learning battery used under each of the plurality of second usage conditions, and the corrected deterioration amount for each correction period of the learning battery as a response variable; further comprising:

9. The information processing method according to claim 8. 　 An information processing device including a processor, The processor: acquiring history information including an accumulated elapsed time for each predetermined period of a prediction target battery used under each of a plurality of first usage conditions expressed by a combination of a predetermined range of SOC and a predetermined range of temperature, and an accumulated charge / discharge amount for each predetermined period of the prediction target battery used under each of a plurality of second usage conditions expressed by a combination of a predetermined range of SOC and a predetermined range of current rate; predicting the deterioration amount of the prediction target battery for each predetermined period by inputting the acquired history information into a learning model; Information processing device. 　 acquiring history information including an accumulated elapsed time for each predetermined period of a prediction target battery used under each of a plurality of first usage conditions expressed by a combination of a predetermined range of SOC and a predetermined range of temperature, and an accumulated charge / discharge amount for each predetermined period of the prediction target battery used under each of a plurality of second usage conditions expressed by a combination of a predetermined range of SOC and a predetermined range of current rate; inputting the acquired history information into a learning model to cause a computer to function so as to predict the deterioration amount of the prediction target battery for each predetermined period; Information processing program.

Images