Battery life prediction device and battery life prediction method

Abstract

This battery life prediction device comprises: an acquisition unit (131) that acquires input data including the temperature of a tire, the pressure of the tire, a value relating to the voltage of a battery, and the usage time of the battery; and a prediction unit (133) that uses the input data and a prediction model to predict, as the life of the battery, a value relating to the voltage after a prescribed period in the future, wherein, where a basic statistic, a cumulative value, a change per unit time, a time-series feature amount, and a travel time with regard to the temperature of the tire and the pressure of the tire in actual results data serve as pre-processing amounts, the prediction model is generated by machine learning by using training data in which at least some of the pre-processing amounts and the value relating to the voltage of the battery serve as explanatory variables and the value relating to the voltage of the battery after the prescribed period serve as a response variable.

Description

Battery life prediction device and battery life prediction method

[0001] The present disclosure relates to a battery life prediction device and a battery life prediction method.

[0002] Various techniques for predicting battery life have been proposed. For example, Patent Document 1 discloses a technique for determining the remaining life of a storage battery for an electric vehicle. In the technique of Patent Document 1, a regression equation is determined by performing a regression analysis using the results of multiple sets of accumulated driving counts and corresponding voltage values. From the regression equation, the number of driving counts corresponding to a life voltage value at a predetermined constant discharge amount is determined, and the current number of driving counts is subtracted to estimate the remaining life driving count.

[0003] Japanese Patent Application Publication No. 06-163084

[0004] Therefore, there is a need for technology to predict the lifespan of batteries used in devices installed in vehicles other than electric vehicles. For example, the sensors and transmitters used in tire pressure monitoring systems (TPMS) are often battery-powered. Predicting battery life and replacing batteries at the appropriate time to prevent TPMS data transmission, which is necessary for tire condition management, is important from the perspective of safe vehicle operation. The technology disclosed in Patent Document 1 estimates the number of full charge cycles remaining until the full charge capacity reaches the limit for use as an electric vehicle battery, because repeated charging and discharging of the battery deteriorates the battery and reduces its full charge capacity. In other words, the technology disclosed in Patent Document 1 predicts changes in full charge capacity due to repeated charging and discharging. However, because the target of prediction is different, it cannot be applied to predicting the battery life (remaining charge) of a TPMS.

[0005] In view of the above circumstances, an object of the present disclosure is to provide a battery life prediction device and a battery life prediction method that can predict with high accuracy the battery life of a detection device that detects information regarding the condition of a tire.

[0006] (1) A battery life prediction device according to one embodiment of the present disclosure is a battery life prediction device that predicts the life of a battery of a detection device that detects information related to the state of a tire, and includes: an acquisition unit that acquires input data including the tire temperature, the tire pressure, values ​​related to the battery voltage, and a usage time of the battery; and a prediction unit that predicts a value related to the voltage after a predetermined period in the future as the life of the battery using the input data and a prediction model, wherein the prediction model is generated by machine learning using learning data in which basic statistics, cumulative values, changes per unit time, time-series features, and running time for the tire temperature and the tire pressure in actual data are preprocessed quantities, the value related to the battery voltage and at least a portion of the preprocessed quantities are explanatory variables, and the value related to the battery voltage after the predetermined period is a target variable.

[0007] (2) As one embodiment of the present disclosure, in (1), the input data is acquired from the detection device by fixed-point observation using a reading device installed at a predetermined location, and the acquisition unit acquires the input data via the reading device.

[0008] (3) As one embodiment of the present disclosure, in (1) or (2), the voltage values ​​are divided into groups of voltage values, and at least one group corresponds to a state of the battery to be replaced.

[0009] (4) As one embodiment of the present disclosure, in (3), the usage time of the battery is the elapsed time since the value related to the voltage first changed.

[0010] (5) As an embodiment of the present disclosure, in any one of (1) to (4), the performance data includes pre-driving data and post-driving data determined based on a change in the temperature of the tire, and the driving time is calculated as the difference between the usage time of the battery included in the pre-driving data and the usage time of the battery included in the post-driving data.

[0011] (6) As an embodiment of the present disclosure, in (5), the cumulative value is calculated as a sum of values ​​obtained by multiplying a difference between the pre-driving data and the post-driving data by a time difference for the tire temperature and the tire pressure, and the time series feature is a trend component or an autoregressive component in a time series analysis of the tire temperature and the tire pressure.

[0012] (7) As an embodiment of the present disclosure, in (5) or (6), the tire temperatures in the pre-travel data are updated based on weather information.

[0013] (8) As an embodiment of the present disclosure, in any one of (1) to (7), the learning data also uses axle position, vehicle ID, fleet ID, and country information as explanatory variables.

[0014] (9) A battery life prediction method according to one embodiment of the present disclosure is a battery life prediction method for predicting the life of a battery of a detection device that detects information related to the state of a tire, the method including: acquiring input data including the tire temperature, the tire pressure, values ​​related to the battery voltage, and a usage time of the battery; and predicting a value related to the voltage after a predetermined period in the future as the life of the battery using the input data and a prediction model, wherein the prediction model is generated by machine learning using learning data in which basic statistics, cumulative values, changes per unit time, time-series features, and running time for the tire temperature and the tire pressure in actual data are preprocessed quantities, the value related to the battery voltage and at least a portion of the preprocessed quantities are explanatory variables, and the value related to the battery voltage after the predetermined period is a target variable.

[0015] According to the present disclosure, it is possible to provide a battery life prediction device and a battery life prediction method that can predict with high accuracy the battery life of a detection device that detects information related to the condition of a tire.

[0016] Fig. 1 is a diagram showing an example configuration of a battery life prediction system including a battery life prediction device according to an embodiment of the present disclosure. Fig. 2 is another diagram showing an example configuration of a battery life prediction system including a battery life prediction device according to an embodiment of the present disclosure. Fig. 3 is a flowchart illustrating a battery life prediction process. Fig. 4 is a flowchart showing details of the model generation step of Fig. 3. Fig. 5 is a flowchart showing details of the battery life prediction step of Fig. 3. Fig. 6 is a diagram for explaining preprocessing.

[0017] A battery life prediction device 10 (see FIG. 1 ) and a battery life prediction method according to an embodiment of the present disclosure will be described below with reference to the drawings. In each drawing, identical or corresponding parts are designated by the same reference numerals. In the description of this embodiment, the description of identical or corresponding parts will be omitted or simplified as appropriate. Here, the "life" of a battery refers to, in the case of a primary battery such as a button battery, alkaline battery, or manganese battery, a state in which the voltage drops to the point where the detection device 70 (see FIG. 1 ) cannot perform its normal function and replacement is required. The functions of the detection device 70 include, for example, a detection function and a communication function. Furthermore, in the case of a secondary battery, the "life" of a battery refers to a state in which the voltage drops to the point where the detection device 70 cannot perform its normal function and charging or replacement is required.

[0018] 1 and 2 are diagrams showing an example of the configuration of a battery life prediction system including a battery life prediction device 10 according to this embodiment. Fig. 1 is a block diagram showing an example of the internal configuration of the battery life prediction device 10. Fig. 2 shows the overall configuration of the battery life prediction system.

[0019] The battery life prediction device 10 predicts the battery life of a detection device 70 that detects information regarding the condition of a tire 30 mounted on a vehicle 20. In this embodiment, the detection device 70 is an on-board device mounted on the vehicle 20 or the tire 30, and specific examples will be described later. The vehicle 20 may be, for example, a passenger car, a truck, a bus, or a construction vehicle, and is not limited to a specific type of moving body. In this embodiment, the vehicle 20 is described as a truck that performs work in a yard (Yard in FIG. 2 ) or that transports goods into or out of the yard. A plurality of vehicles 20 perform work, transport goods into, or transport goods out of the yard.

[0020] In this embodiment, the battery life prediction device 10 generates a prediction model and uses the generated prediction model to predict the battery life of the detection device 70. In other words, the battery life prediction device 10 according to this embodiment is described as a device that has the functions of both a device that generates a prediction model (model generation device) and a device that estimates battery life. As another example, the battery life prediction device 10 may be a device that only has the function of estimating battery life using a prediction model that has been generated in advance.

[0021] As shown in FIG. 1 , the battery life prediction device 10 includes a communication unit 11, a storage unit 12, and a control unit 13. The control unit 13 includes an acquisition unit 131, a model generation unit 132, a prediction unit 133, and an output unit 134. The battery life prediction device 10 may be configured, for example, as a computer in terms of hardware. Details of the components of the battery life prediction device 10 will be described later. Here, the battery life prediction device 10 may not be a single device, but may be configured as a plurality of devices located in multiple locations and capable of transmitting and receiving data to and from each other via a network 40. In other words, the plurality of devices connected via the network 40 may function as the battery life prediction device 10 shown in FIG. 1 as a whole. Therefore, for example, the battery life prediction device 10 may be configured as a single computer in terms of hardware, or may be configured as a plurality of computers connected via the network 40. When configured as a plurality of computers, the storage unit 12 may be a shared memory accessible by each computer or a storage device on the network 40.

[0022] The battery life prediction device 10 may constitute a battery life prediction system together with devices other than the battery life prediction device 10 (hereinafter referred to as "external devices") connected via a network 40. The network 40 is, for example, the Internet. The network 40 may also be configured to include, for example, a local area network (LAN). The external device includes a reading device 60. The external device may also include a terminal device 50 used by a user. The user is a person who uses the battery life prediction system and may include, for example, an administrator who manages maintenance of the vehicle 20. The terminal device 50 is, for example, a general-purpose mobile terminal such as a smartphone or a tablet terminal, but is not limited to these. The external device may also be configured to include an information providing device connected via the network 40. The information providing device may provide, for example, weather information for the area in which the vehicle 20 is traveling. The weather information includes local temperature information and may be used to verify the temperature of the tire 30 (described below) or to correct it to a more accurate value.

[0023] The detection device 70 is a device or an on-board system equipped with a sensor that generates information (hereinafter referred to as "mounted tire information") about the state of the tire 30. In this embodiment, the detection device 70 includes a tire pressure monitoring system (TPMS) that is installed inside the tire 30 or on the wheel.

[0024] The tire pressure monitoring system monitors the pressure (internal pressure) of a tire 30 attached to a vehicle 20. The tire pressure monitoring system may be configured to include, for example, a sensor installed inside the tire 30, a processor that calculates and outputs the pressure of the tire 30 based on the detected value of the sensor, and a memory that stores the detected value of the sensor, etc. The sensor may include a pressure sensor and a temperature sensor.

[0025] In this embodiment, the tire pressure monitoring system outputs information on the pressure and temperature of the tire 30 to the battery life prediction device 10 as the installed tire information. The tire pressure monitoring system is battery-powered. The installed tire information may include not only the pressure and temperature information of the tire 30 but also the battery usage time. For example, if the tire pressure monitoring system has a function for measuring the elapsed time since the first fluctuation in the voltage level (described below), this may be included in the installed tire information as the battery usage time. However, the battery usage time does not have to be measured by the tire pressure monitoring system. For example, at a gate (described below), the elapsed time since the first fluctuation in the voltage level (described below) may be measured through communication between the tire pressure monitoring system and the reading device 60, and this may be used as the battery usage time. As described below, historical data, which is accumulated information on previously installed tires, is used in generating a prediction model. By associating the pressure and temperature information of the tire 30 with time information, the temporal changes in the temperature and pressure of the tire 30 in the historical data can be accurately determined, enabling the generation of a highly accurate prediction model.

[0026] The tire pressure monitoring system may be configured to output the mounted tire information directly to the battery life prediction device 10. However, the tire pressure monitoring system may be powered by a small battery (a battery with a small capacity), such as a button battery. In this embodiment, the tire pressure monitoring system outputs the mounted tire information to the battery life prediction device 10 via the reading device 60 in order to extend the battery life. The tire pressure monitoring system temporarily stores the mounted tire information in memory, and when communication with the reading device 60 becomes possible, transmits the stored mounted tire information to the reading device 60.

[0027] In this embodiment, the tire pressure monitoring system also measures the battery voltage and outputs the battery's "voltage value" to the battery life prediction device 10 as installed tire information. The voltage value is a group of voltage values, and at least one group corresponds to a battery condition requiring replacement. For example, if the voltage at the start of use (i.e., the voltage of a new battery) is 3.1 V, and the measured battery voltage is 3.1 V or higher, the voltage value is "3" (Level 3). If the measured battery voltage is 2.7 V or higher but less than 3.1 V, the voltage value is "2" (Level 2). If the measured battery voltage is 2.4 V or higher but less than 2.7 V, the voltage value is "1" (Level 1). If the measured battery voltage is less than 2.0 V, the voltage value is "0" (Level 0). Of these, Level 1 corresponds to a battery condition requiring replacement. For example, if the voltage value after a predetermined period (e.g., three months) is predicted to be Level 1, the user can avoid the detection device 70 from stopping operation by replacing the battery within the predetermined period. In this way, by using grouped voltage values ​​(levels) instead of the voltage values ​​themselves, efficient battery life prediction becomes possible. Furthermore, the tire pressure monitoring system can transmit battery voltage information with a small amount of data, thereby suppressing an increase in the amount of data transmitted. The above grouping is merely an example, and more groups may be used. Furthermore, the condition of the battery to be replaced may be determined, for example, based on the battery voltage range within which the detection device 70 can transmit installed tire information.

[0028] As shown in FIG. 2 , the reading device 60 is installed, for example, at a gate serving as the entrance and exit to a yard. When the vehicle 20 passes through the gate, the tire pressure monitoring system installed on the vehicle 20 and the reading device 60 are able to communicate with each other, and the installed tire information is read by the reading device 60. In other words, information transmitted from the tire pressure monitoring system is received only when the vehicle 20 approaches the gate. Communication between the tire pressure monitoring system and the reading device 60 may be, for example, short-range wireless communication. The installed tire information is acquired from the detection device 70 by the reading device 60 installed at such a predetermined location through fixed-point observation. The acquisition unit 131 of the battery life prediction device 10 can then acquire the installed tire information via the reading device 60. In the battery life prediction system, the battery of the detection device 70 installed on the vehicle 20 passing through the gate installed in the yard can be used as a prediction target. In other words, the battery life prediction device 10 of this embodiment is capable of highly accurate battery life prediction by the prediction process described below, even for a detection device 70 that does not directly transmit data to the battery life prediction device 10 (a detection device 70 that does not constantly receive the transmitted information).

[0029] The components of the battery life prediction device 10 will be described in detail below. The communication unit 11 includes one or more communication modules connected to the network 40. The communication unit 11 may include a communication module compatible with mobile communication standards such as 4G (4th Generation) and 5G (5th Generation). The communication unit 11 may include a communication module compatible with a wired or wireless LAN standard.

[0030] The storage unit 12 is one or more memories. The memory may be, for example, a semiconductor memory, a magnetic memory, or an optical memory, but is not limited to these, and may be any memory. The storage unit 12 is, for example, built into the battery life prediction device 10, but may also be configured to be accessed from outside by the battery life prediction device 10 via any interface.

[0031] The memory unit 12 stores various data used in the various calculations performed by the control unit 13. The memory unit 12 may also store results and intermediate data of the various calculations performed by the control unit 13. In this embodiment, the memory unit 12 stores the generated prediction model. In this embodiment, the memory unit 12 also stores performance data. The performance data includes past mounted tire information acquired via the reading device 60 and accumulated in the memory unit 12. The performance data includes past tire 30 pressure information, tire 30 temperature information, values ​​related to battery voltage, and battery usage time. In other words, time-series information on past tire 30 pressure, temperature, and voltage values ​​associated with battery usage time is stored in the memory unit 12 as performance data.

[0032] The control unit 13 includes one or more processors. The processor may be, for example, a general-purpose processor or a dedicated processor specialized for a specific process, but is not limited to these and may be any processor. The control unit 13 controls the overall operation of the battery life prediction device 10.

[0033] Here, the battery life prediction device 10 may have the following software configuration: One or more programs used to control the operation of the battery life prediction device 10 are stored in the storage unit 12. When the programs stored in the storage unit 12 are read by the processor of the control unit 13, they cause the control unit 13 to function as an acquisition unit 131, a model generation unit 132, a prediction unit 133, and an output unit 134.

[0034] When generating a prediction model, the acquisition unit 131 acquires historical data for generating learning data. The learning data may be generated by extracting, from the historical data, a target variable (e.g., a value related to the battery voltage after a predetermined period of time) and data for calculating explanatory variables (e.g., the temperature and pressure of the tire 30). When predicting battery life using a prediction model, the acquisition unit 131 acquires the prediction model stored in the storage unit 12. The acquisition unit 131 also acquires input data. The input data includes information on the mounted tire detected by the detection device 70 powered by the battery to be predicted. In other words, the input data includes information on the temperature of the tire 30, the pressure of the tire 30, values ​​related to the battery voltage, and the battery usage time, which are used for battery life prediction.

[0035] The model generation unit 132 generates a prediction model based on the performance data acquired by the acquisition unit 131. The model generation unit 132 performs preprocessing on the tire 30 temperature and tire 30 pressure in the performance data. Preprocessing results in preprocessing quantities, such as basic statistics, cumulative values, changes per unit time, time-series features, and running time. Details of the preprocessing and the preprocessing quantities will be described later. The model generation unit 132 generates training data using the battery voltage value and at least a portion of the preprocessing quantities as explanatory variables and the battery voltage value after a predetermined period as a response variable. In this embodiment, the predetermined period is three months, but may be, for example, six months or one year, and is not limited to a specific value. The model generation unit 132 generates a prediction model through machine learning using the training data. The machine learning method is not limited, but may be, for example, random forest. The number of training data sets is not limited, but may be, for example, 20,000 or more.

[0036] In this embodiment, the model generation unit 132 generates a prediction model without distinguishing between the type or region of the vehicle 20. As another example, the model generation unit 132 may generate a prediction model for each type or region of the vehicle 20. A prediction model tailored to the type or region of the vehicle 20 is selected from multiple prediction models and a battery life prediction is performed, thereby enabling predictions tailored to the circumstances of the type or region of the vehicle 20. For example, battery consumption trends generally differ between cold and warm regions, and predictions that take such circumstances into account are possible. When a prediction model for each type or region of the vehicle 20 is generated, the learning data may also use, for example, axle position, vehicle ID, fleet ID, and country information as explanatory variables. The axle position is the position of the axle on the vehicle 20 and varies depending on the type of vehicle 20. The vehicle ID is an identifier for identifying each vehicle 20. The fleet ID is an identifier given to a group of vehicles 20, such as multiple vehicles 20 working in a yard, and varies depending on the region (work location). The country information is, for example, an identifier for a country and corresponds to regional information for a large division. This information may be added by at least one of the detection device 70 and the reading device 60 and accumulated as performance data.

[0037] As described above, the detection device 70 is an in-vehicle device, and its battery life is affected by factors such as the driving state of the vehicle 20. Therefore, by performing preprocessing on the information included in the performance data so that the driving state of the vehicle 20 is taken into consideration, and by using at least a portion of the values ​​obtained by the preprocessing (preprocessing amount) as explanatory variables for the prediction model, it is possible to generate a prediction model with even higher accuracy. In this embodiment, the model generation unit 132 performs the preprocessing described below.

[0038] FIG. 6 is a diagram for explaining preprocessing. Preprocessing is a process for calculating basic statistics, cumulative values, changes per unit time, time-series feature quantities, and running time, which are preprocessing quantities. In FIG. 6, the temperature of the tire 30 in the actual data is shown in the upper diagram, and the pressure of the tire 30 in the actual data is shown in the lower diagram, as time-series information (line graph). In FIG. 6, the vertical axis in the upper diagram is temperature T, and the vertical axis in the lower diagram is pressure P. The horizontal axis in FIG. 6 is a common time axis.

[0039] In this embodiment, the model generation unit 132 first calculates basic statistics as preprocessing for the temperature and pressure of the tire 30. In this example, the basic statistics are local maximum and minimum values. However, the basic statistics are not limited to local maximum and minimum values, and may include, for example, an average value, a variance, a maximum value, a minimum value, etc. In the example of FIG. 6 , at time t ｉ＋１ The maximum value of temperature T at ｉ＋１ and the maximum value P of pressure P ｉ＋１ In the example of FIG. 6, at time t ｉ The minimum value of temperature T at ｉ and the minimum value P of the pressure P ｉ In this way, the model generating unit 132 obtains the maximum and minimum values ​​as basic statistics for the entire time-series information of the temperature and pressure of the tire 30.

[0040] The model generation unit 132 determines whether the tire temperature T is pre-driving data or post-driving data of the performance data based on the change in the tire temperature T of the tire 30. When the vehicle 20 drives, the tire temperature T of the tire 30 rises due to friction heat with the road surface caused by the vehicle driving. In the example of FIG. 6, the maximum value T ｉ＋１ is the post-driving data for temperature T. And at the same time t ｉ＋１ The maximum value P ｉ＋１ is the post-travel data for the pressure P. In the example of FIG. 6, the minimum value T ｉ is the data before the trip for the temperature T. ｉ The minimum value P ｉ is the pre-driving data for the pressure P. In this way, the actual data includes pre-driving data and post-driving data determined based on the change in temperature of the tire 30. The model generation unit 132 can further calculate the driving time. The driving time is calculated as the difference between the battery usage time included in the pre-driving data and the battery usage time included in the post-driving data. This calculation makes it possible to associate the actual data with the driving state, thereby enabling the generation of a more accurate prediction model that takes the driving state into account.

[0041] The model generation unit 132 may use basic statistics as explanatory variables of the learning data, but it is more preferable to use cumulative values ​​based on basic statistics, changes per unit time, time-series feature quantities, and running time as explanatory variables. The cumulative values ​​are calculated as the sum of values ​​obtained by multiplying the difference between pre-run data and post-run data by the time difference for the temperature and pressure of the tire 30. For example, the cumulative value for temperature T is calculated by the following formula (1). Also, for example, the cumulative value for pressure P is calculated by the following formula (2). However, in this embodiment, formula (1) is used to calculate T ｉ＋１ >T ｉ Similarly, equation (2) is P ｉ＋１ >P ｉ Calculations were performed only if

[0042]

[0043] The change per unit time is calculated by dividing the difference between the pre-travel data and the post-travel data by the time difference. For example, the change per unit time for the temperature T is calculated using the following formula (3). Also, the change per unit time for the pressure P is calculated using the following formula (4). However, in this embodiment, formula (3) is used to calculate the temperature T. ｉ＋１ >T ｉ Similarly, equation (4) is used to calculate the increase after driving only when P ｉ＋１ >P ｉ Calculations were performed only if

[0044]

[0045] The time-series feature values ​​may be trend components or autoregressive components in a time-series analysis of the tire temperature and pressure. These parameters based on basic statistics are also associated with the driving state, enabling the generation of a more accurate prediction model that takes the driving state into account. The model generation unit 132 may further perform processing such as normalization on the parameters used as explanatory variables. The model generation unit 132 may also perform the above-described preprocessing after interpolating actual data obtained through sparse fixed-point observations using an interpolation technique between data. The model generation unit 132 may also update the tire temperature 30 in the pre-driving data based on weather information. Because the tire temperature 30 before driving the vehicle 20 is considered to be the same as the air temperature, overwriting the pre-driving tire temperature 30 with the weather information data allows the tire temperature 30 before driving to be corrected to a more accurate value, enabling the generation of a more accurate prediction model.

[0046] The model generation unit 132 may generate multiple prediction models (candidate prediction models) using multiple machine learning techniques using training data in which the battery voltage value and at least a portion of the preprocessing amount are used as explanatory variables and the battery voltage value after a predetermined period is used as a target variable. In this case, the model generation unit 132 may select one prediction model from the candidate prediction models based on an index. The multiple machine learning techniques may include, for example, random forest, convolutional neural network, regression neural network, etc. The index used to evaluate the generated candidate prediction models may be precision, accuracy, recall, specificity, F-measure, or a combination of at least some of these.

[0047] The prediction unit 133 uses the input data acquired by the acquisition unit 131 and the prediction model stored in the storage unit 12 to predict a voltage value at a predetermined time in the future as the battery life. As described above, the prediction model is generated by machine learning using learning data based on past time-series information on tire 30 pressure and temperature associated with the battery's usage time. In particular, with a gated TPMS, tire 30 pressure and temperature data is acquired through fixed-point observation, and only intermittent, discontinuous tire 30 pressure and temperature data may be obtained. However, by performing machine learning using accumulated performance data, a highly accurate prediction model is generated. Therefore, the battery life of the detection device 70, which detects information on the condition of the tire 30, can be predicted with high accuracy.

[0048] In this embodiment, the explanatory variables of the prediction model are the above-mentioned preprocessing amounts, etc. The prediction unit 133 performs preprocessing on the input data in the same manner as the model generation unit 132. Then, the prediction unit 133 inputs the preprocessing amounts obtained by the preprocessing into the prediction model.

[0049] The output unit 134 outputs the results predicted by the prediction unit 133 to a display device or the like. For example, a display included in the terminal device 50 can function as the display device that displays the prediction results. The prediction results include values ​​related to the predicted voltage (e.g., the voltage level after a predetermined period of time) and are provided to the user via the network 40. The user can access predicted information about the battery life of the detection device 70 wherever they are. For example, the user can plan battery replacement work for the detection device 70 based on the prediction results. For example, the prediction unit 133 may periodically perform predictions for the detection devices 70 of multiple vehicles 20 operating in a yard. The output unit 134 may periodically display the prediction results to the user, thereby preventing oversight of battery replacement work. Furthermore, the accuracy of the prediction results may be periodically verified based on actual measurements. For example, if the accuracy rate falls below a predetermined reference value, the model generation unit 132 may regenerate or update the prediction model using performance data added after the generation of the prediction model. Here, performance data may also be used to verify the accuracy of the prediction results. For example, the performance data may be divided into training data (learning data), validation data, test data, etc., and used for different purposes.

[0050] 3 is a flowchart illustrating the battery life prediction process executed by the battery life prediction device 10. The acquisition unit 131 and the model generation unit 132 execute a model generation step (step S1, model generation method) to generate a prediction model for predicting battery life. The acquisition unit 131, the prediction unit 133, and the output unit 134 execute a battery life prediction step (step S2, battery life prediction method) to predict the life of the battery that is the target of prediction using the generated prediction model. The battery life prediction step may be executed consecutively with the model generation step, or may be executed a certain time after the model generation step.

[0051] Fig. 4 is a flowchart showing the details of the model generation process of Fig. 3. The acquisition unit 131 acquires performance data stored in, for example, the storage unit 12 (step S11).

[0052] The model generation unit 132 executes the above preprocessing and calculates the amount of preprocessing (step S12). The model generation unit 132 generates training data using the calculated amount of preprocessing (step S13), and generates a prediction model by machine learning using the training data (step S14). The generated prediction model is stored in the storage unit 12.

[0053] Fig. 5 is a flowchart showing the details of the battery life prediction process of Fig. 3. The acquisition unit 131 acquires input data and performs preprocessing (step S21), and acquires a prediction model from the storage unit 12 (step S22).

[0054] The prediction unit 133 uses the input data and the prediction model to predict a voltage value at a predetermined time in the future as the battery life (step S23). The output unit 134 outputs the result predicted by the prediction unit 133 (step S24). The output prediction result may be displayed on the display of the terminal device 50, for example, so that an administrator who manages maintenance of the vehicles 20 can check it. For example, if the prediction result predicts that the batteries of the detection devices 70 installed in 10 vehicles 20 will reach Level 1 (a battery condition requiring replacement) in three months, the administrator may schedule battery replacement for those 10 vehicles 20 within three months.

[0055] As described above, the battery life prediction device 10 and battery life prediction method according to this embodiment, with the above-described configuration and process, are capable of highly accurately predicting the battery life of the detection device 70 that detects information about the condition of the tire 30. When the method of this embodiment was verified, it was found that when the change per unit time was included in the explanatory variables, the prediction accuracy improved by 1% to 2% compared to when it was not included. Furthermore, when the cumulative value was included in the explanatory variables, the prediction accuracy improved by 5% to 6% compared to when it was not included.

[0056] Although the embodiments of the present disclosure have been described based on the drawings and examples, it should be noted that those skilled in the art would easily be able to make various modifications or alterations based on the present disclosure. Therefore, it should be noted that these modifications and alterations are included within the scope of the present disclosure. For example, the functions included in each component or step can be rearranged so as not to cause logical inconsistencies, and multiple components or steps can be combined or divided into one. The embodiments of the present disclosure can also be realized as a program executed by a processor included in an apparatus and a storage medium on which the program is recorded. It should be understood that these are also included within the scope of the present disclosure.

[0057] For example, the configurations of the battery life prediction device 10 and the battery life prediction system shown in FIGS. 1 and 2 are merely examples and are not limited to the configurations shown in FIGS. 1 and 2 . For example, the battery life prediction system may be configured with the battery life prediction device 10 and the reading device 60 integrated together. Furthermore, in FIG. 2 , weather information does not need to be provided from the information providing device. Furthermore, the memory unit 12 may not be a memory device included in the battery life prediction device 10, but may be configured as a memory device on the network 40 (a cloud-based memory device) as seen from the battery life prediction device 10. For example, performance data may be accumulated in a memory device on the network 40. Furthermore, values ​​related to the prediction model and the predicted voltage (prediction results) may be stored in a memory device on the network 40. With this configuration, the battery life prediction device 10 is not limited to a device with a large memory device (e.g., a server computer), but can be configured as, for example, a small computer (e.g., a tablet terminal).

[0058] Furthermore, the model generation unit 132 and the prediction unit 133 may be included in different computers. For example, the model generation unit 132 may be included in another computer (an independent model generation device) that can communicate with the battery life prediction device 10 and can also access the storage unit 12. In this case, the battery life prediction device 10 may predict battery life using a prediction model that is generated by the other computer and stored in the storage unit 12. The prediction model may be generated in the same manner as described above. In other words, the other computer may acquire performance data and generate a prediction model using learning data based on the performance data.

[0059] REFERENCE SIGNS LIST 10 Battery life prediction device 11 Communication unit 12 Storage unit 13 Control unit 20 Vehicle 30 Tire 40 Network 50 Terminal device 60 Reading device 70 Detection device 131 Acquisition unit 132 Model generation unit 133 Prediction unit 134 Output unit

Claims

1. A battery life prediction device that predicts the life of a battery of a detection device that detects information regarding the condition of a tire, comprising: an acquisition unit that acquires input data including the tire temperature, the tire pressure, values ​​related to the battery voltage, and the usage time of the battery; and a prediction unit that uses the input data and a prediction model to predict a value related to the voltage after a predetermined period of time in the future as the life of the battery, wherein the prediction model is generated by machine learning using learning data in which basic statistics, cumulative values, changes per unit time, time series features, and driving time for the tire temperature and the tire pressure in actual data are preprocessed quantities, the value related to the battery voltage and at least a portion of the preprocessed quantities are explanatory variables, and the value related to the battery voltage after the predetermined period is an objective variable.

2. A battery life prediction device as described in claim 1, wherein the input data is obtained from the detection device by fixed-point observation using a reading device installed at a predetermined location, and the acquisition unit acquires the input data via the reading device.

3. A battery life prediction device according to claim 1 or 2, wherein the voltage-related values ​​are divided into groups of voltage values, and at least one group corresponds to a state in which the battery should be replaced.

4. A battery life prediction device according to claim 3, wherein the battery usage time is the time elapsed since the voltage-related value first fluctuated.

5. A battery life prediction device as described in any one of claims 1 to 4, wherein the performance data includes pre-driving data and post-driving data determined based on changes in tire temperature, and the driving time is calculated as the difference between the usage time of the battery included in the pre-driving data and the usage time of the battery included in the post-driving data.

6. The battery life prediction device of claim 5, wherein the cumulative value is calculated as the sum of values ​​obtained by multiplying the difference between the pre-driving data and the post-driving data for the tire temperature and the tire pressure by a time difference, and the time series feature is a trend component or an autoregressive component in a time series analysis of the tire temperature and the tire pressure.

7. The battery life prediction device according to claim 5 or 6, wherein the tire temperature of the pre-travel data is updated based on meteorological information.

8. A battery life prediction device according to any one of claims 1 to 7, wherein the learning data also uses axle position, vehicle ID, fleet ID and country information as explanatory variables.

9. A battery life prediction method for predicting the life of a battery of a detection device that detects information regarding the condition of a tire, comprising: acquiring input data including the tire temperature, the tire pressure, values ​​related to the battery voltage, and a usage time of the battery; and predicting a value related to the voltage after a predetermined period of time in the future as the life of the battery using the input data and a prediction model, wherein the prediction model is generated by machine learning using learning data in which basic statistics, cumulative values, changes per unit time, time-series features, and driving time for the tire temperature and the tire pressure in actual data are preprocessed quantities, the value related to the battery voltage and at least a portion of the preprocessed quantities are explanatory variables, and the value related to the battery voltage after the predetermined period is an objective variable.

Images