Device, system, method, and program for predicting evaluation result, etc. of storage battery

Abstract

The evaluation result prediction device, system, method, and program according to the present disclosure are configured to: extract a plurality of explanatory variables from data items pertaining to a lifespan evaluation test, which are test condition data items and charge / discharge data items, or from charge / discharge data items including an energy density and a power density; input the plurality of explanatory variables to a trained model; and thereby be able to predict a prediction evaluation result measured by a predetermined lifespan evaluation test or an energy density and a power density measured by a predetermined charge / discharge protocol. An operable factor prediction device, system, method, and program according to the present disclosure calculate a plurality of deterioration levels corresponding to lifespan determination events and a plurality of deterioration events of a storage battery from the prediction evaluation result, and thereby can predict, as operable factors, one or a plurality of operable factor candidates having a causal relationship with the lifespan determination events, from an association model that associates in advance the lifespan determination events and the plurality of deterioration events with the plurality of operable factor candidates to identify causal relationships therebetween.

Description

Apparatus, system, method, and program for predicting battery evaluation results, etc.

[0001] The present disclosure relates to an apparatus, system, method, and program for predicting the evaluation results of a storage battery, and a method, apparatus, system, and program for predicting operable factors that affect the life of a storage battery.

[0002] In recent years, the development of storage batteries, typified by lithium-ion secondary batteries, has been vigorously pursued. Storage batteries are widely used in power storage systems in the fields of vehicles, communications, industry, renewable energy, and electrical and electronic devices. The performance of storage batteries is evaluated in various situations depending on the development stage and product lifecycle by various entities that develop and / or use storage batteries, such as storage battery material manufacturers, storage battery manufacturers, primary users of storage batteries (e.g., product manufacturers), secondary users (e.g., product users), and recyclers.

[0003] As the development of storage batteries accelerates, it is rational for developers and users of storage batteries to outsource performance evaluation and analysis of storage batteries to battery evaluation contractors, rather than owning assets such as evaluation facilities. Battery evaluation contractors (hereinafter simply referred to as "contractors") are required to evaluate and analyze the performance of storage batteries entrusted to them by various business entities (hereinafter simply referred to as "customers") and provide accurate information as feedback quickly.

[0004] Some performance evaluation tests for storage batteries are completed in a short period of time, while others, such as life evaluation tests, require a storage battery to be stored for a certain period of time under specified conditions or to repeatedly undergo charge-discharge cycles under specified conditions, which require a considerable period of time. For example, the storage test for storage batteries specified in IEC 62660-1 (Performance Test), standardized by the International Electrotechnical Commission (IEC), requires measuring the voltage retention rate after storing the storage battery at 45°C for 28 days and measuring the capacity retention rate three times after storing the storage battery at 45°C for 42 days. Furthermore, the cycle test specified in IEC 62660-1 requires repeating a specified charge-discharge cycle at 45°C over 28 days until one of the following conditions is met: (i) six cycles are completed (168 days of evaluation is completed), (ii) the capacity retention rate falls below 80%, or (iii) the temperature of the storage battery reaches the upper limit temperature.

[0005] Furthermore, some battery performance evaluation tests require a considerable amount of time due to the need for multiple charge / discharge steps and adjustment of test conditions. For example, a test to measure the initial battery performance, energy density and power density, according to the method specified in IEC 62660-1 (Performance Test), standardized by the International Electrotechnical Commission (IEC), requires a predetermined charge / discharge protocol, including measuring the energy density by discharging the battery at a predetermined temperature and current value (C rate) and measuring the power density under multiple temperature and state of charge (SOC) conditions. Each step of adjusting the battery to the predetermined temperature and SOC takes several hours, and completing the entire charge / discharge protocol takes approximately three days in total.

[0006] On the other hand, there are known techniques for determining the performance and degradation state of a storage battery based on charge / discharge data of the storage battery. For example, Patent Document 1 describes a degradation determination system, a degradation determination method, and a degradation determination program that determine the SOH (State of Health) of a storage battery based on acquired data of the storage battery, perform linear regression on time-series data of the degradation of the storage battery, and determine whether or not the storage battery has suddenly deteriorated. Patent Document 2 describes a method for determining the SOH of a storage battery using a characteristic charge / discharge protocol.

[0007] In light of the above background, the present inventors have recently succeeded in developing a technology that can predict the evaluation results of a storage battery's life evaluation test and shorten the time required for the test. This technology uses data science to combine battery charge / discharge data, electrochemical analysis data, and material analysis data that reflect the battery's degradation events, thereby enabling predicted deterioration information for the storage battery after degradation to be obtained even with limited data, achieving both prediction accuracy and interpretability. The predicted deterioration information after degradation may also include the degree of deterioration of the battery's components, such as the positive electrode, negative electrode, separator, and electrolyte. This has the advantage of allowing users to understand the cause of battery degradation, consider the degradation mechanism, and plan countermeasures against degradation. However, simply obtaining predicted deterioration information after degradation does not determine which specific storage battery design should be manipulated and how to efficiently improve the battery's lifespan.

[0008] International Publication No. WO 2023 / 026743 International Publication No. WO 2023 / 084010

[0009] In one embodiment, one object of the present disclosure is to provide an evaluation result prediction device, system, method, and program that can predict the evaluation results of a life evaluation test of a storage battery and reduce the time required for the life evaluation test of a storage battery.

[0010] In one embodiment, an object of the present disclosure is to provide an evaluation result prediction device, system, method, and program that can predict the evaluation results of the energy density and power density of a storage battery and can shorten the time required for performance evaluation tests of the energy density and power density of the storage battery.

[0011] In one embodiment, an object of the present disclosure is to provide a method, an apparatus, a system, and a program that can predict operable factors that affect the lifespan of a storage battery, thereby contributing to efficiently improving the lifespan of the storage battery.

[0012] Examples of the first embodiment of the present disclosure are listed in the following items [1] to

[24] . [1] An evaluation result prediction device that predicts at least one evaluation result selected from a post-test voltage retention rate, a post-test capacity retention rate, a time to reach life, or a number of cycles to reach life of a storage battery, which are measured in a predetermined life evaluation test, wherein the evaluation result prediction device comprises: an acquisition unit that acquires at least one test condition data selected from a temperature, a current value, a voltage, or a charge / discharge rest time of the storage battery, and charge / discharge data of the storage battery based on an evaluation charge / discharge protocol including measuring energy density and power density during the life evaluation test; an explanatory variable extraction unit that extracts a plurality of explanatory variables including one or more test condition parameters selected from the test condition data and electrochemical parameters including the energy density and the power density calculated from the charge / discharge data; a storage unit that stores a trained model that has been trained to output a predicted evaluation result including the evaluation result measured in the life evaluation test when the plurality of explanatory variables are input; and an evaluation prediction unit that acquires the predicted evaluation result by inputting the plurality of explanatory variables into the trained model, and calculates the prediction accuracy of the predicted evaluation result. and an output unit that outputs information including the predicted evaluation result and the prediction accuracy. [2] The evaluation result prediction device according to item 1, wherein the predetermined life evaluation test includes: a degradation step in which a storage battery is stored at a predetermined temperature for a fixed period (X) or a fixed number (Y) of charge / discharge cycles are repeated at a predetermined temperature, and an evaluation step that includes (i) measuring a voltage retention rate and / or a capacity retention rate of the storage battery after completion of the degradation step, or (ii) measuring a life reaching time or a number of life reaching cycles until the voltage retention rate or capacity retention rate of the storage battery falls below a predetermined reference value or until the storage battery reaches a predetermined upper limit temperature.[3] The evaluation charge / discharge protocol includes measuring the energy density and the power density in a single charge / discharge process, and the single charge / discharge process includes discharging the storage battery at the predetermined temperature and a predetermined current value to measure the energy density, and temporarily discharging the storage battery to a maximum current value I when the charging rate reaches a target charging rate during the discharging process. ｍａｘand measuring the power density at the target charging rate, and then returning the current value to the predetermined value again, each time one or more target charging rates are reached, thereby measuring the energy density and the power density in a single discharge process. [4] The evaluation result prediction device according to item 2 or 3, wherein the evaluation charge / discharge protocol includes a plurality of single charge / discharge processes in which the energy density and the power density are measured by a single charge / discharge process at regular intervals shorter than X or at regular intervals of a certain number of charge / discharge cycles less than Y during the life evaluation test; and the evaluation result prediction device is configured to obtain the predicted evaluation result based on the test condition data and charge / discharge data of the single charge / discharge process and calculate the prediction accuracy; and the evaluation result prediction device further includes a control unit, and the control unit controls the evaluation result prediction device to: (i) terminate the life evaluation test if the prediction accuracy is equal to or greater than a predetermined threshold; and (ii) continue the life evaluation test and perform a next single charge / discharge process in the evaluation charge / discharge protocol to obtain further charge / discharge data, and repeat the process of obtaining the predicted evaluation result and calculating the prediction accuracy based on the test condition data and accumulated charge / discharge data, until the prediction accuracy becomes equal to or greater than the threshold. [5] The evaluation result prediction device according to any one of items 1 to 4, wherein the electrochemical parameters further include a capacity retention rate of the storage battery. [6] The evaluation result prediction device according to any one of items 1 to 5, wherein the acquisition unit is configured to further acquire at least one material analysis data selected from an X-ray diffraction (XRD) spectrum, a nuclear magnetic resonance (NMR) spectrum, an electron spin resonance (ESR) spectrum, a scanning electron microscope (SEM) image, or an ion chromatography (IC) spectrum of a material constituting the storage battery, and the plurality of explanatory variables further include one or more material analysis parameters obtained by mathematically processing the material analysis data in addition to the test condition parameters and the electrochemical parameters.[7] The evaluation result prediction device according to any one of items 1 to 6, wherein the plurality of explanatory variables further include, in addition to the test condition parameters and the electrochemical parameters, one or more mathematical parameters obtained by mathematically processing the shape of the discharge curve in the single charge / discharge process. [8] The evaluation result prediction device according to any one of items 1 to 7, wherein the plurality of explanatory variables are explanatory variables pre-selected by excluding one or more parameters that are multicollinear with other parameters so as to improve the prediction accuracy of the trained model or improve interpretability of the predicted evaluation result. [9] The evaluation result prediction device according to item 7, wherein the mathematical parameters are explanatory variables pre-selected from C coefficients in matrix W obtained by matrix decomposing, with mathematical constraints, a 1-row, B-column matrix X in which voltage (V) values ​​corresponding to B (B≧2) different state of charge (SOC) values ​​are arranged as elements in the column direction, into X≈W×H (where matrix W is a 1-row, C-column coefficient matrix, and matrix H is a C-row, B-column basis matrix).

[10] The evaluation result prediction device according to item 6, wherein the predicted evaluation result further includes predicted deterioration information of the storage battery when it reaches the end of its life.

[11] The evaluation result prediction device according to any one of items 1 to 10, wherein the output unit is configured to output information including the predicted evaluation result and the prediction accuracy to a virtual environment accessible by a user terminal.

[12] The evaluation result prediction device according to item 11, further configured to terminate the evaluation charge / discharge protocol when a command to terminate the evaluation charge / discharge protocol is received from the virtual environment.

[13] An evaluation result prediction system that predicts at least one evaluation result selected from a post-test voltage retention rate, a post-test capacity retention rate, a time to reach life, or a number of cycles to reach life of a storage battery, which are measured in a predetermined life evaluation test, wherein the evaluation result prediction system comprises: an acquisition device that acquires at least one test condition data selected from a temperature, a current value, a voltage, or a charge / discharge rest time of the storage battery, and charge / discharge data of the storage battery based on an evaluation charge / discharge protocol including measuring energy density and power density during the life evaluation test; an explanatory variable extraction device that extracts a plurality of explanatory variables including one or more test condition parameters selected from the test condition data and electrochemical parameters including the energy density and the power density calculated from the charge / discharge data; a storage device that stores a trained model that is trained to output a predicted evaluation result including the evaluation result measured in the life evaluation test when the plurality of explanatory variables are input; and an evaluation prediction device that acquires the predicted evaluation result by inputting the plurality of explanatory variables into the trained model, and calculates the prediction accuracy of the predicted evaluation result. and an output device that outputs information including the predicted evaluation result and the prediction accuracy.

[14] The evaluation result prediction system according to item 13, wherein the predetermined life evaluation test includes: a degradation step in which a storage battery is stored at a predetermined temperature for a fixed period (X) or a fixed number (Y) of charge / discharge cycles are repeated at a predetermined temperature; and an evaluation step that includes: (i) measuring a voltage retention rate and / or a capacity retention rate of the storage battery after completion of the degradation step, or (ii) measuring a life reaching time or a number of life reaching cycles until the voltage retention rate or capacity retention rate of the storage battery falls below a predetermined reference value or the storage battery reaches a predetermined upper limit temperature.

[15] The evaluation charge / discharge protocol includes measuring the energy density and the power density in a single charge / discharge process, the single charge / discharge process including: measuring the energy density by discharging the storage battery at the predetermined temperature and a predetermined current value; and temporarily setting a maximum current value I when a charging rate reaches a target charging rate during the discharging process.ｍａｘand measuring the power density at the target charging rate, and then returning the current to the predetermined current value again, each time one or more target charging rates are reached, thereby measuring the energy density and the power density in a single discharge process.

[16] The evaluation result prediction system according to item 14 or 15, wherein the evaluation charge / discharge protocol includes a plurality of single charge / discharge processes, in which measurements of the energy density and the power density are performed by a single charge / discharge process at regular intervals shorter than X or at regular intervals of a certain number of charge / discharge cycles less than Y during the life evaluation test; and the evaluation result prediction system is configured to obtain the predicted evaluation result based on the test condition data and charge / discharge data of the single charge / discharge processes, and calculate the prediction accuracy; and the evaluation result prediction system further includes a control device, which controls the evaluation result prediction system to: (i) terminate the life evaluation test if the prediction accuracy is equal to or greater than a predetermined threshold; and (ii) continue the life evaluation test and perform a next single charge / discharge process in the evaluation charge / discharge protocol to obtain further charge / discharge data, and obtain the predicted evaluation result and calculate the prediction accuracy based on the test condition data and accumulated charge / discharge data, until the prediction accuracy becomes equal to or greater than the threshold.

[17] An evaluation result prediction method for predicting at least one evaluation result selected from a post-test voltage retention rate, a post-test capacity retention rate, a time to reach life, or a number of cycles to reach life of a storage battery, which are measured in a predetermined life evaluation test, the evaluation result prediction method comprising: acquiring at least one test condition data selected from a temperature, a current value, a voltage, or a charge / discharge rest time of the storage battery, and charge / discharge data of the storage battery based on an evaluation charge / discharge protocol including measuring energy density and power density during the life evaluation test; extracting a plurality of explanatory variables including one or more test condition parameters selected from the test condition data and electrochemical parameters including the energy density and the power density calculated from the charge / discharge data; acquiring the predicted evaluation result by inputting the plurality of explanatory variables into a trained model that is trained to output a predicted evaluation result including the evaluation result measured in the life evaluation test when the plurality of explanatory variables are input; calculating a prediction accuracy of the predicted evaluation result; and outputting information including the predicted evaluation result and the prediction accuracy.

[18] The method for predicting an evaluation result according to Item 17, wherein the predetermined life evaluation test includes: a degradation step of storing a storage battery at a predetermined temperature for a certain period (X) or repeating a certain number (Y) of charge-discharge cycles at a predetermined temperature; and an evaluation step including: (i) measuring a voltage retention rate and / or a capacity retention rate of the storage battery after completion of the degradation step, or (ii) measuring a life reaching time or a number of life reaching cycles until the voltage retention rate or capacity retention rate of the storage battery falls below a predetermined reference value or the storage battery reaches a predetermined upper limit temperature.

[19] The evaluation charge-discharge protocol includes measuring the energy density and the power density in a single charge-discharge process, the single charge-discharge process including: discharging the storage battery at the predetermined temperature and a predetermined current value to measure the energy density; and during the discharging process, temporarily increasing a maximum current value I when a charge rate reaches a target charge rate. ｍａｘand measuring the power density at the target charging rate, and then returning the current to the predetermined value again, each time one or more target charging rates are reached, thereby measuring the energy density and the power density in a single discharge process.

[20] The evaluation charge / discharge protocol includes a plurality of single charge / discharge processes in which the energy density and the power density are measured by a single charge / discharge process at regular intervals shorter than X or at regular intervals of a certain number of charge / discharge cycles less than Y during the life evaluation test, and the evaluation result prediction method further includes: acquiring the predicted evaluation result based on the test condition data and charge / discharge data of the single charge / discharge process and calculating the prediction accuracy; (i) if the prediction accuracy is equal to or greater than a predetermined threshold, terminating the life evaluation test; and (ii) if the prediction accuracy is less than the threshold, continuing the life evaluation test and performing a next single charge / discharge process in the evaluation charge / discharge protocol to acquire further charge / discharge data, and repeating the steps of acquiring the predicted evaluation result and calculating the prediction accuracy based on the test condition data and accumulated charge / discharge data until the prediction accuracy becomes equal to or greater than the threshold.

[21] An evaluation result prediction program that predicts at least one evaluation result selected from a post-test voltage retention rate, a post-test capacity retention rate, a time to reach life, or a number of cycles to reach life of a storage battery, which are measured in a predetermined life evaluation test, wherein the evaluation result prediction program causes a computer to execute the following steps: acquire at least one test condition data selected from a temperature, a current value, a voltage, or a charge / discharge rest time of the storage battery, and charge / discharge data of the storage battery based on an evaluation charge / discharge protocol that includes measuring energy density and power density during the life evaluation test; extract a plurality of explanatory variables including one or more test condition parameters selected from the test condition data and electrochemical parameters including the energy density and the power density calculated from the charge / discharge data; acquire the predicted evaluation result by inputting the plurality of explanatory variables into a trained model that is trained to output a predicted evaluation result including the evaluation result measured in the life evaluation test when the plurality of explanatory variables are input; calculate a prediction accuracy of the predicted evaluation result; and output information including the predicted evaluation result and the prediction accuracy.

[22] The predetermined life evaluation test includes: a degradation step of storing a storage battery at a predetermined temperature for a certain period (X) or repeating a certain number (Y) of charge-discharge cycles at a predetermined temperature; and an evaluation step including: (i) measuring a voltage retention rate and / or a capacity retention rate of the storage battery after completion of the degradation step, or (ii) measuring a life reaching time or a number of life reaching cycles until the voltage retention rate or capacity retention rate of the storage battery falls below a predetermined reference value or the storage battery reaches a predetermined upper limit temperature.

[23] The evaluation charge-discharge protocol includes measuring the energy density and the power density in a single charge-discharge process, the single charge-discharge process including: discharging the storage battery at the predetermined temperature and a predetermined current value to measure the energy density; and during the discharging process, temporarily increasing a maximum current value I when the charging rate reaches the target charging rate. ｍａｘand measuring the power density at the target charging rate, and then returning the current value to the predetermined value again, each time one or more target charging rates are reached, thereby measuring the energy density and the power density in a single discharge process.

[24] The evaluation result prediction program according to item 22 or 23, further causing a computer to execute the following: the evaluation charge / discharge protocol includes a plurality of single charge / discharge processes, in which measurements of the energy density and the power density are performed by a single charge / discharge process at regular intervals shorter than X or at regular intervals of a certain number of charge / discharge cycles less than Y during the life evaluation test; and the evaluation result prediction program further causes a computer to execute the following: (i) obtain the predicted evaluation result based on the test condition data and charge / discharge data of the single charge / discharge process, and calculate the prediction accuracy; and (ii) if the prediction accuracy is less than the threshold, continue the life evaluation test and perform a next single charge / discharge process in the evaluation charge / discharge protocol to obtain further charge / discharge data, and obtain the predicted evaluation result and calculate the prediction accuracy based on the test condition data and accumulated charge / discharge data, until the prediction accuracy becomes equal to or greater than the threshold.

[0013] Examples of the second embodiment of the present disclosure are listed in the following items [1] to

[21] . [1] An evaluation result prediction device that predicts the energy density and power density of a storage battery that are sequentially measured using a predetermined charge / discharge protocol, the evaluation result prediction device comprising: an acquisition unit that acquires charge / discharge data of the storage battery based on an evaluation charge / discharge protocol that includes measuring the energy density and the power density in a single charge / discharge process; an explanatory variable extraction unit that extracts, from the charge / discharge data, a plurality of explanatory variables including electrochemical parameters including the energy density and the power density; a storage unit that stores a trained model that is trained to output a predicted evaluation result including the energy density and the power density measured using the predetermined charge / discharge protocol when the plurality of explanatory variables are input; an evaluation prediction unit that acquires the predicted evaluation result by inputting the plurality of explanatory variables into the trained model and calculates a prediction accuracy of the predicted evaluation result; and an output unit that outputs information including the predicted evaluation result and the prediction accuracy. [2] The predetermined charge / discharge protocol includes an energy density measurement step of discharging a storage battery at a first temperature and a first current value to measure the energy density, and a charge or discharge process of adjusting the storage battery to a target charge rate and achieving a maximum current value I ｍａｘ and one or more power density measurement steps, wherein the single charge / discharge process comprises: discharging the storage battery at the first temperature and at a second current value greater than the first current value to measure an energy density; and temporarily increasing a maximum current value I when the charging rate reaches the target charging rate during the discharging process. ｍａｘand measuring the power density at the target charging rate, and then returning the current to the second current value again, each time one or more of the target charging rates are reached, thereby measuring the energy density and the power density in a single discharge process. [4] The evaluation charge / discharge protocol includes first to n-th single charge / discharge processes, each of which measures energy density and power density at first to n-th different temperatures (n is an integer of 2 or more) in a single charge / discharge process, and the first to n-th single charge / discharge processes are configured in an order in which the prediction accuracy becomes higher or the temperature adjustment is completed in a shorter time; the evaluation result prediction device is configured to obtain the predicted evaluation result based on charge / discharge data of a first single charge / discharge process and calculate the prediction accuracy; the evaluation result prediction device further includes a control unit, and the control unit: (i) terminates the evaluation charge / discharge protocol if the prediction accuracy is equal to or higher than a predetermined threshold; and (ii) performs a next single charge / discharge process in the evaluation charge / discharge protocol, obtains further charge / discharge data, and repeats obtaining the predicted evaluation result and calculating the prediction accuracy based on the accumulated charge / discharge data, until the prediction accuracy becomes equal to or higher than the threshold, or until the n-th single charge / discharge process is completed. [5] The evaluation result prediction device according to any one of items 1 to 3, wherein the plurality of explanatory variables further include, in addition to the electrochemical parameters, one or more mathematical parameters obtained by mathematically processing the shape of the discharge curve in the single charge / discharge process. [6] The evaluation result prediction device according to item 5, wherein the plurality of explanatory variables are explanatory variables selected in advance by excluding one or more parameters that are multicollinear with other parameters so as to improve the prediction accuracy of the learning model or improve interpretability of the predicted evaluation result.[7] The evaluation result prediction device according to item 6, wherein the mathematical parameters are explanatory variables pre-selected from C coefficients in matrix W obtained by matrix decomposing a 1-row, B-column matrix X in which voltage (V) values ​​corresponding to B (B≧2) different state of charge (SOC) values ​​are arranged as elements in the column direction for discharge curve data in the single charge / discharge process, into X≈W×H (wherein matrix W is a 1-row, C-column coefficient matrix, and matrix H is a C-row, B-column basis matrix) after adding mathematical constraints. [8] The evaluation result prediction device according to any one of items 1 to 7, wherein the output unit is configured to output information including the predicted evaluation result and the prediction accuracy to a virtual environment accessible by a user terminal. [9] The evaluation result prediction device according to item 8, further configured to terminate the evaluation charge / discharge protocol when a command to terminate the evaluation charge / discharge protocol is received from the virtual environment.

[10] An evaluation result prediction system that predicts the energy density and power density of a storage battery that are measured sequentially using a predetermined charge / discharge protocol, the evaluation result prediction system comprising: an acquisition device that acquires charge / discharge data of the storage battery based on an evaluation charge / discharge protocol that includes measuring the energy density and power density in a single charge / discharge process; an explanatory variable extraction device that extracts, from the charge / discharge data, a plurality of explanatory variables including electrochemical parameters including the energy density and the power density; a storage device that stores a trained model that is trained to output a predicted evaluation result including the energy density and power density measured using the predetermined charge / discharge protocol when the plurality of explanatory variables are input; an evaluation prediction device that acquires the predicted evaluation result by inputting the plurality of explanatory variables into the trained model and calculates the prediction accuracy of the predicted evaluation result; and an output device that outputs information including the predicted evaluation result and the prediction accuracy.

[11] The predetermined charge / discharge protocol includes an energy density measurement step of discharging a storage battery at a first temperature and a first current value to measure its energy density; and adjusting the storage battery to a target charge rate by a charge or discharge process that is different from the energy density measurement step in temperature, charge or discharge, or current value, or is different in two or more of these, and achieving a maximum current value I. ｍａｘ and one or more power density measurement steps, wherein the single charge / discharge process comprises: discharging the storage battery at the first temperature and at a second current value greater than the first current value to measure an energy density; and temporarily increasing the maximum current value I when the charging rate reaches the target charging rate during the discharging process. ｍａｘand measuring a power density at the target charging rate, and then returning the current to the second current value again, each time one or more target charging rates are reached, thereby measuring the energy density and the power density in a single discharge process.

[13] The evaluation charge / discharge protocol includes first to n-th single charge / discharge processes in which energy density and power density are measured at first to n-th different temperatures (n is an integer of 2 or more) in a single charge / discharge process, and the first to n-th single charge / discharge processes are configured in an order in which the prediction accuracy becomes higher or the temperature adjustment is completed in a shorter time; the evaluation result prediction system is configured to obtain the predicted evaluation result based on charge / discharge data of a first single charge / discharge process and calculate the prediction accuracy; and the evaluation result prediction system further includes a control device, which: (i) terminates the evaluation charge / discharge protocol if the prediction accuracy is equal to or higher than a predetermined threshold; and (ii) performs a next single charge / discharge process in the evaluation charge / discharge protocol, obtains further charge / discharge data, and repeats obtaining the predicted evaluation result and calculating the prediction accuracy based on the accumulated charge / discharge data, until the prediction accuracy becomes equal to or higher than the threshold, or until the n-th single charge / discharge process is completed. 13. The evaluation result prediction system according to any one of items 10 to 12, wherein the evaluation result prediction system is controlled as follows:

[14] A method for predicting an energy density and a power density of a storage battery that are sequentially measured using a predetermined charge / discharge protocol, the method comprising: acquiring charge / discharge data of the storage battery based on an evaluation charge / discharge protocol that includes measuring the energy density and the power density in a single charge / discharge process; extracting from the charge / discharge data a plurality of explanatory variables including electrochemical parameters including the energy density and the power density; acquiring the predicted evaluation result by inputting the plurality of explanatory variables into a trained model that is trained to output a predicted evaluation result including the energy density and the power density measured using the predetermined charge / discharge protocol when the plurality of explanatory variables are input; calculating a prediction accuracy of the predicted evaluation result; and outputting information including the predicted evaluation result and the prediction accuracy.

[15] The predetermined charge / discharge protocol includes: an energy density measurement step of discharging a storage battery at a first temperature and a first current value to measure its energy density; and adjusting the storage battery to a target charge rate by a charge or discharge process that is different from the energy density measurement step in temperature, charge or discharge, or current value, or two or more of these are different, and a maximum current value I is reached. ｍａｘ and one or more power density measurement steps, wherein the single charge / discharge process comprises: discharging the storage battery at the first temperature and at a second current value greater than the first current value to measure an energy density; and temporarily increasing a maximum current value I when the charging rate reaches the target charging rate during the discharging process. ｍａｘand measuring a power density at the target charging rate, and then returning the current to the second current value again, each time one or more target charging rates are reached, thereby measuring the energy density and the power density in a single discharge process.

[17] The evaluation charge / discharge protocol includes first to n-th single charge / discharge processes in which energy density and power density are measured in a single charge / discharge process at first to n-th different temperatures (n is an integer of 2 or more), respectively, and the first to n-th single charge / discharge processes are configured in an order in which the prediction accuracy becomes higher or the temperature adjustment is completed in a shorter time; and the evaluation result prediction method further includes: acquiring the predicted evaluation result based on charge / discharge data of a first single charge / discharge process and calculating the prediction accuracy; (i) if the prediction accuracy is equal to or higher than a predetermined threshold, terminating the evaluation charge / discharge protocol; and (ii) if the prediction accuracy is less than the threshold, performing a next single charge / discharge process in the evaluation charge / discharge protocol, acquiring further charge / discharge data, and acquiring the predicted evaluation result and calculating the prediction accuracy based on the accumulated charge / discharge data; and repeating these steps until the prediction accuracy becomes equal to or higher than the threshold or until the n-th single charge / discharge process is completed.

[18] An evaluation result prediction program that predicts the energy density and power density of a storage battery that are sequentially measured using a predetermined charge / discharge protocol, the evaluation result prediction program causing a computer to execute the following steps: acquire charge / discharge data of the storage battery based on an evaluation charge / discharge protocol that includes measuring the energy density and power density in a single charge / discharge process; extract from the charge / discharge data a plurality of explanatory variables including electrochemical parameters including the energy density and the power density; acquire the predicted evaluation result by inputting the plurality of explanatory variables into a trained model that is trained to output a predicted evaluation result including the energy density and power density measured using the predetermined charge / discharge protocol when the plurality of explanatory variables are input; calculate a prediction accuracy of the predicted evaluation result; and output information including the predicted evaluation result and the prediction accuracy.

[19] The predetermined charge / discharge protocol includes an energy density measurement step of discharging a storage battery at a first temperature and a first current value to measure its energy density; and adjusting the storage battery to a target charge rate by a charge or discharge process that is different from the energy density measurement step in temperature, charge or discharge, or current value, or two or more of these are different, and a maximum current value I is reached. ｍａｘ and one or more power density measurement steps, wherein the single charge / discharge process comprises: discharging the storage battery at the first temperature and at a second current value greater than the first current value to measure an energy density; and temporarily increasing a maximum current value I when the charging rate reaches the target charging rate during the discharging process. ｍａｘand measuring a power density at the target charging rate, and then returning the current to the second current value again, each time one or more target charging rates are reached, thereby measuring the energy density and the power density in a single discharge process.

[21] The evaluation charge / discharge protocol includes first to n-th single charge / discharge processes in which energy density and power density are measured in a single charge / discharge process at first to n-th different temperatures (n is an integer of 2 or more), respectively, and the first to n-th single charge / discharge processes are configured in an order in which the prediction accuracy becomes higher or the temperature adjustment is completed in a shorter time; and the evaluation result prediction program further causes a computer to execute the following: (i) acquire the predicted evaluation result based on charge / discharge data of a first single charge / discharge process and calculate the prediction accuracy; and (ii) if the prediction accuracy is less than the threshold, perform a next single charge / discharge process in the evaluation charge / discharge protocol, acquire further charge / discharge data, and acquire the predicted evaluation result and calculate the prediction accuracy based on the accumulated charge / discharge data, until the prediction accuracy becomes equal to or higher than the threshold or until the n-th single charge / discharge process is completed.

[0014] Examples of the third embodiment of the present disclosure are listed in the following items [1] to

[12] . [1] A method for predicting operable factors that affect the life of a storage battery, the method comprising: a step of calculating a life determining event that is a basis for determining the life of the storage battery and a plurality of deterioration levels that indicate the degree of deterioration, each corresponding to a plurality of deterioration events that are directly or indirectly causally related to the life determining event, from at least one piece of predicted deterioration information selected from the group consisting of predicted charge / discharge characteristics, predicted electrochemical analysis data, and predicted material analysis data after deterioration of the storage battery; a step of simplifying an association model that previously associates the causal relationships between the life determining event, the plurality of deterioration events, and a plurality of operable factor candidates by excluding one or more deterioration events whose deterioration level is lower than a threshold, and as a result, excluding any deterioration events and operable factor candidates whose causal relationship with the life determining event has been severed; and a step of outputting one or more operable factor candidates that are directly or indirectly causally related to the life determining event in the simplified association model as operable factors that affect the life of the storage battery. [2] The method further includes acquiring the influence degrees corresponding to one or more of the plurality of degradation events by inputting the deterioration degree of the lifespan determining event and the deterioration degrees of the plurality of degradation events into a trained model that is trained to output, when the deterioration degree of the lifespan determining event and the deterioration degrees of the plurality of degradation events are input, an influence degree that represents an improvement effect on the deterioration degree of the lifespan determining event when one or more of the plurality of degradation events are improved, and outputting, in the simplified association model, one or more of the candidate operable factors that have a direct or indirect causal relationship with one or more of the degradation events whose influence degree is equal to or greater than a threshold, as an operable factor affecting the lifespan of the storage battery.[3] The method for predicting an operational factor according to item 2, further comprising: correlating the operational factor with an input and / or output of the trained model using a theoretical calculation formula expressing the relationship between the operational factor and the deterioration levels and / or impact levels of the plurality of degradation events in the simplified association model, and predicting a change in the lifespan of the storage battery when each of the output operational factors is changed by performing forward analysis and / or reverse analysis of the trained model from a specific operational factor. [4] The method for predicting an operational factor according to item 2, further comprising: constructing a simulation model simulating a change in the lifespan of the storage battery when each of the output operational factors is changed using a theoretical calculation formula expressing the relationship between the output operational factor, the deterioration levels of the plurality of degradation events, and the deterioration level of a lifespan determining event in the simplified association model, and simulating a change in the lifespan of the storage battery when each of the output operational factors is changed using the simulation model, and outputting a simulation result. [5] The method for predicting operable factors according to any one of items 2 to 4, wherein the life determining event is capacity degradation of the storage battery. [6] The method for predicting operable factors according to any one of items 2 to 5, wherein the association model is a tree-shaped association model based on fault tree analysis (FTA) with the life determining event as a top event, and is configured with the plurality of degradation events as upper layers and the plurality of candidate operable factors as lower layers. [7] The method for predicting operable factors according to item 6, wherein the plurality of degradation events include, from highest to lowest layers, one or more charge / discharge characteristic degradation events, multiple electrochemical analysis degradation events, and multiple material analysis degradation events, and the plurality of candidate operable factors are located in lower layers than each of the multiple material analysis degradation events.[8] The method for predicting an operable factor according to any one of items 1 to 7, wherein the predicted deterioration information includes predicted charge / discharge characteristics, predicted electrochemical analysis data, and predicted material analysis data of the storage battery at the end of its life, the predicted charge / discharge characteristics include at least one selected from the group consisting of energy density, power density, capacity, and charge / discharge efficiency of the storage battery at the end of its life, the predicted electrochemical analysis data includes at least one selected from the group consisting of charge curve analysis (CCA), discharge curve analysis, constant current intermittent titration (GITT), and electrochemical impedance spectroscopy (EIS) of the storage battery at the end of its life, and the predicted material analysis data includes at least one selected from the group consisting of nuclear magnetic resonance (NMR) data, ion chromatography (IC) data, scanning electron microscope (SEM) observation data, X-ray diffraction (XRD) data, electron spin resonance (ESR) data, X-ray CT image data, and air permeability measurement data of materials constituting the storage battery at the end of its life. [9] The operable factor candidates include the density, thickness, volume resistivity, interface resistance, peel strength, porosity and basis weight of the negative electrode active material layer, the type of negative electrode active material, the particle structure, particle size, specific surface area, crystal structure, crystallinity, the amount and content of impurities of the negative electrode active material, the type of negative electrode binder and its content, silicon oxide (SiOx) content, the density, thickness, volume resistivity, interface resistance, peel strength, porosity and basis weight of the positive electrode active material layer, the type of positive electrode active material, the particle structure, particle size, specific surface area, crystal structure, crystallinity, the amount and content of impurities of the positive electrode active material, the type of positive electrode binder and its content, and the material of the conductive additive. the type and content of the filler in the coating layer, the particle size and content of the filler in the coating layer, the type and content of the binder in the coating layer, and the coating position of the coating layer (single-sided coating on the positive electrode side, single-sided coating on the negative electrode side, or double-sided coating).

[10] An operationable factor prediction device for predicting operationable factors that affect the life of a storage battery, the device comprising: an acquisition unit that acquires at least one piece of predicted deterioration information selected from the group consisting of predicted charge / discharge characteristics, predicted electrochemical analysis data, and predicted material analysis data after deterioration of the storage battery; a deterioration degree calculation unit that calculates, from the predicted deterioration information, a life determining event that is a criterion for determining the life of the storage battery, and a plurality of deterioration levels that indicate the degree of deterioration, each corresponding to a plurality of deterioration events that are directly or indirectly causally related to the life determining event; a storage unit that stores an association model that previously associates the causal relationships between the life determining event, the plurality of deterioration events, and a plurality of operable factor candidates; and a simplified association model generation unit that simplifies the association model by excluding from the association model one or a plurality of deterioration events whose deterioration level is lower than a threshold, and as a result, excluding any of the deterioration events and the operable factor candidates whose causal relationship with the life determining event has been severed; an output unit that outputs one or more candidate operable factors that have a direct or indirect causal relationship with the life-determining event in the simplified correlation model as operable factors that affect the life of the storage battery.

[11] A system for predicting operable factors that affect the life of a storage battery, comprising: an acquisition device that acquires at least one piece of predicted deterioration information selected from the group consisting of predicted charge / discharge characteristics, predicted electrochemical analysis data, and predicted material analysis data after deterioration of the storage battery; a deterioration degree calculation device that calculates, from the predicted deterioration information, a life determining event that is a criterion for determining the life of the storage battery, and a plurality of deterioration levels that indicate the degree of deterioration, each corresponding to a plurality of deterioration events that are directly or indirectly causally related to the life determining event; a storage device that stores an association model that previously associates the causal relationships between the life determining event, the plurality of deterioration events, and a plurality of candidate operable factors; and a simplified association model generation device that simplifies the association model by excluding from the association model one or more deterioration events whose deterioration level is lower than a threshold, and as a result, excluding any deterioration events and candidate operable factors that have lost their causal relationship with the life determining event; an output device that outputs one or more candidate operable factors that have a direct or indirect causal relationship with the life-determining event in the simplified correlation model as operable factors that affect the life of the storage battery.

[12] An operable factor prediction program for predicting operable factors that affect the life of a storage battery, the program comprising: calculating a life determining event that is a criterion for determining the life of the storage battery and a plurality of deterioration levels that indicate the degree of deterioration, each corresponding to a plurality of deterioration events that are directly or indirectly causally related to the life determining event, from at least one predicted deterioration information selected from the group consisting of predicted charge / discharge characteristics, predicted electrochemical analysis data, and predicted material analysis data after deterioration of the storage battery; simplifying an association model that previously associates the causal relationships between the life determining event, the plurality of deterioration events, and a plurality of operable factor candidates by excluding one or more deterioration events whose deterioration level is lower than a threshold, and as a result, excluding any deterioration events and operable factor candidates whose causal relationship with the life determining event has been severed; and outputting one or more operable factor candidates that are directly or indirectly causally related to the life determining event in the simplified association model as operable factors that affect the life of the storage battery. A program for predicting operable factors, which is characterized by causing a computer to execute the above.

[0015] According to one embodiment of the present disclosure, an evaluation result prediction device, system, method, and program are provided that can predict the evaluation results of a life evaluation test of a storage battery and can shorten the time required for the life evaluation test of a storage battery.

[0016] According to one embodiment of the present disclosure, there is provided an evaluation result prediction device, system, method, and program that can predict the evaluation results of the energy density and power density of a storage battery and can shorten the time required for performance evaluation testing of the energy density and power density of the storage battery.

[0017] According to one embodiment of the present disclosure, a method, an apparatus, a system, and a program are provided that can predict operational factors that affect the lifespan of a storage battery, thereby contributing to efficiently improving the lifespan of the storage battery.

[0018] FIG. 1 is a diagram showing a schematic configuration of an evaluation result prediction system 1 according to an embodiment of the present disclosure. FIG. 2(a) is a diagram showing an example of a single discharge process in an evaluation charge / discharge protocol. FIG. 2(b) is a diagram showing an example of an evaluation charge / discharge protocol including performing a single discharge process every certain number of charge / discharge cycles. FIG. 3 is a diagram showing an example of the evaluation result prediction system 1 according to the embodiment of the present disclosure in FIG. 1. FIG. 4 is a diagram showing an example of calculation of mathematical parameters as explanatory variables. FIG. 5 is a flowchart showing an example of operation of a learning process of a trained model. FIG. 6 is a flowchart showing an example of operation of a learning process of a trained model. FIG. 7 is a flowchart showing an example of operation of an evaluation prediction process (an evaluation result prediction method) using a trained model obtained by the learning process of FIG. 5. FIG. 8 is a flowchart showing an example of operation of an evaluation prediction process (an evaluation result prediction method) using a trained model obtained by the learning process of FIG. 6. FIG. 9 is a diagram showing an example of the evaluation result prediction system 1 according to the embodiment of the present disclosure in FIG. 1. FIG. 10 is a diagram showing a schematic configuration of the evaluation result prediction system 1 according to an embodiment of the present disclosure. FIG. 11 is a diagram showing an example of the evaluation result prediction system 1 according to the embodiment of the present disclosure in FIG. 10. FIG. 12 is a flowchart showing an example of the operation of the learning process for a trained model. FIG. 13 is a flowchart showing an example of the operation of the evaluation prediction process (evaluation result prediction method). FIG. 14 is a flowchart showing an example of the operation ...FIG. 24 is a diagram showing a schematic configuration of the evaluation result and manipulable factor prediction system of the present disclosure.

[0019] Hereinafter, embodiments of the present disclosure will be described in detail, but the technical scope of the present invention is not limited to the following embodiments and encompasses the inventions described in the claims and their equivalents. The upper and lower limits of each numerical range in the following embodiments can be arbitrarily combined to form any numerical range.

[0020] <Evaluation Result Prediction System and Evaluation Result Prediction Device> FIG. 1 is a diagram illustrating a schematic configuration of an evaluation result prediction system 1 according to an embodiment of the present disclosure. The evaluation result prediction system 1 includes a charging / discharging device 100, a charging / discharging device terminal 200 communicatively connected to the charging / discharging device 100, a business operator server 300 communicatively connected to the charging / discharging device terminal 200 via a business operator's private network N1, and a virtual environment 400 constructed on a network N2 and communicatively connected to the business operator server 300. The evaluation result prediction system 1 is accessible from terminal devices, such as a business operator terminal 500 and a customer terminal 600, which are indicated by dashed lines. The business operator terminal 500 is communicatively connected to the charging / discharging device terminal 200 and the business operator server 300 via the business operator's private network N1, and is communicatively connected to the virtual environment 400 and the customer terminal 600 via the network N2. The customer terminal 600 is communicatively connected to the virtual environment 400 and the business operator terminal 500 via the network N2. The evaluation result prediction system 1 can predict the evaluation result that would be obtained if a predetermined life evaluation test were performed on the storage battery to be evaluated, and deterioration information of the storage battery at the end of its life, based on test condition data in the life evaluation test and charge / discharge data obtained by charging and discharging the storage battery to be evaluated based on an evaluation charge / discharge protocol during the life evaluation test.In place of or in addition to predicting the evaluation result of the predetermined life evaluation test, the evaluation result prediction system 1 can predict the evaluation result that would be obtained if a predetermined charge / discharge protocol were performed on the storage battery to be evaluated, based on charge / discharge data obtained by charging and discharging the storage battery to be evaluated based on the evaluation charge / discharge protocol.

[0021] The evaluation result prediction system 1 may be configured to predict, in addition to the evaluation results of the predetermined life evaluation test, and preferably, in addition to the evaluation results of the predetermined life evaluation test and the evaluation results according to the predetermined charge / discharge protocol, an operable factor, which is a design factor of the storage battery that affects the life of the storage battery. In this case, the evaluation result prediction system 1 may include an operable factor prediction device 800 (not shown), as exemplified in FIGS. 11 and 12 . For example, at least one information processing device selected from the group consisting of the charging / discharging device terminal 200, the business operator server 300, and the shared server 700 (if present) in the evaluation result prediction system 1 may further include a configuration as the operable factor prediction device 800. Alternatively, the evaluation result prediction system 1 may further include, in another information processing device communicatively connected to one or more of the components via the network N1 and / or N2, a configuration as the operable factor prediction device 800. Alternatively, the configuration of the operable factor prediction system may be distributed so that the charging / discharging device terminal 200, the business operator server 300, the shared server 700 (if present), and the other information processing device (if present) function as a whole as an operable factor prediction system.

[0022] <Predetermined Life Evaluation Test> The life evaluation test is a test that includes degrading a storage battery under predetermined conditions for a long period of time. The predetermined life evaluation test includes measuring at least one evaluation result selected from a post-test voltage retention rate, a post-test capacity retention rate, a life time, a life cycle count, or a degradation state at the end of life of the storage battery, which are to be predicted by the evaluation result prediction system and evaluation result prediction device of the present disclosure.

[0023] The predetermined life evaluation test includes, for example, a degradation step in which the storage battery is stored at a predetermined temperature for a certain period (X) or a certain number (Y) of charge / discharge cycles are repeated at a predetermined temperature, and an evaluation step in which the evaluation results are measured. The evaluation step includes, for example, (i) measuring the voltage retention rate, capacity retention rate, or both of the storage battery after the degradation step is completed. This makes it possible to obtain the post-test voltage retention rate, post-test capacity retention rate, or both of the storage battery as evaluation results. Instead of or in addition to (i), the evaluation step includes (ii) measuring the life time or the number of life cycles until the voltage retention rate or capacity retention rate of the storage battery falls below a predetermined reference value or until the storage battery reaches a predetermined upper limit temperature. This makes it possible to obtain the life time, the number of life cycles, or both of the storage battery as evaluation results.

[0024] More specifically, examples of the predetermined life evaluation test include a storage battery storage test and a cycle test specified in IEC 62660-1 (Performance Test) standardized by the International Electrotechnical Commission (IEC). The storage battery storage test in IEC 62660-1 includes a test to measure a voltage retention rate and a test to measure a capacity retention rate. The test to measure the voltage retention rate includes a degradation step in which the storage battery is adjusted to a state of charge (SOC) of 50% under predetermined conditions and stored at 45°C for 28 days (X = 28 days), and an evaluation step in which the voltage retention rate after the degradation step is measured. The test to measure the capacity retention rate includes a degradation step in which the storage battery is adjusted to a state of charge (SOC) of 100% or 50% under predetermined conditions and stored at 45°C for 42 days, and an evaluation step in which the capacity retention rate of the storage battery is measured after the degradation step, repeated three times (X = 126 days). Furthermore, the cycle test for a storage battery in IEC 62660-1 includes a degradation step in which a predetermined charge / discharge cycle is repeated at 45°C over 28 days, and an evaluation step in which the temperature and capacity retention rate of the storage battery are measured after the degradation step. The number of charge / discharge cycles (Y) repeated over 28 days is determined by the charge / discharge conditions of the predetermined charge / discharge cycle. For example, if charging / discharging is performed at a current value of 1 C with a charge / discharge rest period of 1 hour, each cycle takes 3 hours, resulting in 224 cycles (Y = 224) over 28 days. The degradation step and evaluation step are repeated until one or more of the following conditions are met (X = 28 to 168 days, Y = 224 to 1344): (i) the degradation step is completed a total of six times, (ii) the capacity retention rate of the storage battery reaches 80%, or (iii) the temperature of the storage battery reaches a predetermined upper limit temperature.

[0025] The predetermined charge / discharge protocol is a charge / discharge protocol for measuring the energy density and / or power density to be predicted by the evaluation result prediction system and evaluation result prediction device of the present disclosure, and includes sequentially measuring the energy density and / or power density. The predetermined charge / discharge protocol is composed of multiple measurement steps that differ in temperature, whether the process is a charge process or a discharge process, or the current value (C rate), or that differ in two or more of these.

[0026] The predetermined charge / discharge protocol may include, for example, an energy density measurement step of discharging the storage battery at a first temperature and a first current value (C rate) to measure the energy density, and a charge or discharge process different from the energy density measurement step of adjusting the storage battery to a target charge rate and charging the storage battery to a maximum current value I ｍａｘ and one or more power density measurement steps, which sequentially measure the power density at the target charging rate. The energy density measurement step and the one or more power density measurement steps are performed at different charging and discharging steps, which are different in temperature, whether they are charging or discharging processes, or which have different current values ​​(C rates), or which differ in two or more of these. The power density measurement step may include, for example, multiple power density measurement steps, which measure the power density at one or more target charging rates at first to n-th different temperatures (n is an integer of 2 or more).

[0027] More specifically, the predetermined charge / discharge protocol includes a test for measuring energy density and power density, as defined in IEC 62660-1 (Performance Test) standardized by the International Electrotechnical Commission (IEC). When measuring the energy density and power density of a battery based on the method defined in IEC 62660-1, for example, the following steps must be taken: (1) adjusting the storage battery to an SOC of 100% under predetermined conditions, and discharging the storage battery from an SOC of 100% to 0% at 25°C (first temperature) and a current value of 1 / 3 C (first current value), thereby measuring the energy density; and (2) charging the storage battery to an SOC of 20% at 25°C (first temperature), and discharging the storage battery at a maximum current value Iｍａｘ (3) charging the storage battery to an SOC of 50% at 25°C (first temperature) and measuring the power density at a maximum current value I ｍａｘ (4) charging the storage battery to an SOC of 80% at 25°C (first temperature) and measuring the power density at a maximum current value I ｍａｘ (5) discharging the storage battery to an SOC of 50% at −20° C. (a second temperature) and measuring the power density at 25° C. and an SOC of 80%; and ｍａｘ (6) adjusting the storage battery to an SOC of 50% at 0°C (a third temperature) and measuring the power density at the maximum current value I ｍａｘ (7) adjusting the storage battery to an SOC of 50% at 45°C (a fourth temperature) and measuring the power density at the maximum current value I ｍａｘ and measuring the power density at 45° C. and SOC 50% by outputting the voltage for 10 seconds. An example of the above charge / discharge protocol in accordance with IEC 62660-1 is shown in Table 1 below.

[0028]

[0029] Step (1) takes about 3 hours to discharge because the current value is 1 / 3 C, and steps (2) to (7) each take 10 seconds to discharge. However, it takes several hours to adjust the battery to a predetermined temperature and SOC before each step, and it takes about 3 days in total to complete the entire charge / discharge protocol.

[0030] <Evaluation Charging / Discharging Protocol> The evaluation charging / discharging protocol is a protocol for charging and discharging a storage battery to be evaluated for prediction processing by the evaluation result prediction system and evaluation result prediction device, and includes a charging / discharging process for measuring energy density and power density. The evaluation result prediction system and evaluation result prediction device acquire not only test condition data for the life evaluation test but also charge / discharge data of the storage battery during the life evaluation test, and use explanatory variables extracted from the test condition data and the charge / discharge data in the evaluation prediction process, thereby enabling rapid and accurate prediction of evaluation results during the life evaluation test. This reduces the time required for the life evaluation test of the storage battery. The evaluation charging / discharging protocol may measure energy density and power density sequentially or in a single charging / discharging process. From the perspective of minimizing the impact on the life evaluation test of the storage battery and shortening the time required for the life evaluation test, it is preferable to measure energy density and power density in a single charging / discharging process.

[0031] The single charge / discharge process can measure the power density when a target charge rate is reached during a discharge process for measuring the energy density, for example. More specifically, the single charge / discharge process may include, for example, measuring the energy density by discharging the storage battery at a predetermined temperature and a predetermined current value (C rate). When predicting evaluation results including the energy density and / or power density according to a predetermined charge / discharge protocol, the single charge / discharge process may include, for example, measuring the energy density by discharging the storage battery at a first temperature that is the same as the first temperature of the predetermined charge / discharge protocol and at a second current value (C rate) that is greater than the first current value (C rate) of the predetermined charge / discharge protocol. Since the second current value (C rate) is greater than the first current value (C rate), the time required for discharge is shortened, thereby further shortening the time required for a performance evaluation test of the storage battery. Then, during the discharge process, when the charge rate reaches the target charge rate, a maximum current value I is temporarily set. ｍａｘThe evaluation charge / discharge protocol may include outputting a current value of 1 / 3 C, measuring the power density at the target charge rate, and then returning the current value to a predetermined value and continuing the discharge. The evaluation charge / discharge protocol may have multiple target charge rates, and the power density may be measured each time the target charge rate is reached. This allows the energy density and one or more power densities to be measured in a single discharge process. The predetermined temperature may be the same temperature as the temperature in the predetermined life evaluation test to be predicted. A large predetermined current value (C rate) shortens the time required for discharge, thereby further shortening the time required for the evaluation and prediction process of the storage battery. The lower limit of the predetermined current value (C rate) may be, for example, 1 / 3 C or more, 1 / 2 C or more, or 1 C or more, and the upper limit may be, for example, 2 C or less, or 1.5 C or less. When predicting evaluation results including energy density and / or power density according to a predetermined charge / discharge protocol, for a single charge / discharge process, for example, the lower limit of the second current value (C rate) can be, for example, 1.5 times or more, 2.0 times or more, 2.5 times or more, 3.0 times or more, or 3.5 times or more of the first current value (C rate), and the upper limit can be, for example, 5.0 times or less, 4.5 times or less, or 4.0 times or less of the first current value (C rate). ｍａｘ The lower limit value can be, for example, 1 C or more, 2 C or more, or 3 C or more, and the upper limit value can be 5 C or less, 4 C or less, or 3 C or less.

[0032] The evaluation charge / discharge protocol preferably includes first to mth (m is an integer of 2 or greater) evaluation charge / discharge protocols, each of which includes measuring energy density and power density in one or more single charge / discharge processes at intervals of a fixed period X' shorter than the period X during which the storage battery is stored in the specified life evaluation test, or at intervals of a fixed number Y' of charge / discharge cycles shorter than the number Y of charge / discharge cycles performed in the specified life evaluation test. When the evaluation charge / discharge protocol includes multiple such evaluation charge / discharge protocols, the evaluation result prediction system and evaluation result prediction device may be configured to obtain predicted evaluation results and calculate prediction accuracy each time an evaluation charge / discharge protocol is performed. The frequency of performing the evaluation charge / discharge protocols can be appropriately set from the perspectives of shortening the time required for the evaluation / prediction process, achieving high prediction accuracy, and not significantly affecting the life evaluation test. The frequency (period) X' of performing the evaluation charge / discharge protocol may be ½ or less, ⅓ or less, ¼ or less, ⅕ or less, ⅙ or less, ⅓ ...

[0033] When the predetermined charge / discharge protocol includes measuring energy density and / or power density at first to n different temperatures (n is an integer of 2 or greater), the evaluation charge / discharge protocol may include first to n single charge / discharge processes, each of which is performed at a corresponding first to n different temperatures (n is an integer of 2 or greater). The order in which the first to n single charge / discharge processes are performed may correspond to the measurement order in the predetermined charge / discharge protocol, or may be any order that does not correspond to the measurement order. For example, the order in which the first to n single charge / discharge processes are performed may be configured to increase the prediction accuracy of the trained model or to complete temperature adjustment in a shorter time. When the evaluation charge / discharge protocol includes first to n single charge / discharge processes, it is preferable to perform all of the first to n single charge / discharge processes from the perspective of achieving higher prediction accuracy. However, from the perspective of shortening the time required for the evaluation prediction process, it is not necessary to perform all of the single charge / discharge processes; it is preferable to terminate the evaluation charge / discharge protocol once the user's desired prediction accuracy is obtained.

[0034] When predicting the evaluation results of a predetermined life evaluation test using a predetermined charge / discharge protocol, the first to mth evaluation charge / discharge protocols may be performed for a certain period or for a certain number of charge / discharge cycles, and the evaluation charge / discharge protocols may each include the first to nth single charge / discharge processes.

[0035] 2A is a diagram showing an example of a single discharge process in the evaluation charge / discharge protocol. The single discharge process in FIG. 2A includes measuring the energy density of a storage battery adjusted to 100% SOC by discharging the storage battery at a current value of 1 C until the SOC drops from 100% to 0% (Evaluation A). In addition, during the process of performing Evaluation A, when the SOC of the storage battery reaches 80%, a maximum current value I of 2 C or more is measured. ｍａｘfor 10 seconds to measure the power density at an SOC of 80%, and then suspend charging and discharging for 1 minute, after which discharging is continued again at a current value of 1 C (Evaluation B). Similar operations are performed when the SOC of the storage battery reaches 50% and when it reaches 20%, and the power densities at SOCs of 50% and 20% are measured (Evaluation C and Evaluation D).

[0036] The evaluation charge / discharge protocol may include first performing the single discharge process illustrated in FIG. 2(a) on a new storage battery at the start of a life evaluation test, and then performing the discharge process every fixed period (X') shorter than the storage period (X) of the storage battery in the specified life evaluation test, or every fixed number of charge / discharge cycles (Y') shorter than the number of charge / discharge cycles (Y) performed in the specified life evaluation test. By performing the single discharge process at the start of the life evaluation test, charge / discharge data of the new storage battery can be obtained, and electrochemical parameters in the new state, such as energy density, power density, and discharge capacity, can be calculated from the obtained charge / discharge data. For example, when the evaluation results of a storage battery storage test (X = 28 days or 42 days) according to IEC 62660-1 are to be predicted, the evaluation charge / discharge protocol may include performing the single discharge process illustrated in FIG. 2(a) at 45°C, for example, every 14 days, every 7 days, every 4 days, every 2 days, or every day. When the evaluation results of a cycle test of a storage battery according to IEC 62660-1 are to be predicted, the evaluation charge / discharge protocol may include performing a single discharge process, as illustrated in FIG. 2(a), at 45°C, for example, every 14 days, 7 days, 4 days, 2 days, or every day. Alternatively, the evaluation charge / discharge protocol may include performing the single discharge process at 45°C, for example, every 100 cycles, 50 cycles, 25 cycles, 20 cycles, or 10 cycles. This allows predicted evaluation results and prediction accuracy to be obtained for a certain period or every certain number of charge / discharge cycles. Once the user's desired prediction accuracy is achieved, the evaluation charge / discharge protocol can be terminated (i.e., the life evaluation test can be terminated), eliminating the need for subsequent life evaluation tests. This significantly reduces the time required for life evaluation testing. For example, FIG. 2(b) shows an example of an evaluation charge / discharge protocol that includes performing a single discharge process every certain number of charge / discharge cycles. FIG. 2(b) shows that the single discharge process of FIG. 2(a) described above is first performed on a new storage battery at the start of a life evaluation test, and then performed every time the number of charge / discharge cycles (e.g., Y′=100) is less than the number of charge / discharge cycles performed in the cycle test of a storage battery in IEC 62660-1.

[0037] The evaluation charge / discharge protocol may include first to n-th single charge / discharge processes, in which the single discharge process illustrated in FIG. 2(a) is performed at first to n-th different temperatures (n is an integer of 2 or greater). For example, when the evaluation results of IEC 62660-1 are to be predicted as the predetermined charge / discharge protocol, the evaluation charge / discharge protocol includes four single charge / discharge processes, in which the single discharge process illustrated in FIG. 2(a) is performed at 25°C (first temperature), -20°C (second temperature), 0°C (third temperature), and 45°C (fourth temperature). When configuring the order of the charge / discharge processes to increase the prediction accuracy of the trained model, the order may be, for example, a first single charge / discharge process at 25°C (first temperature), a second single charge / discharge process at -20°C (second temperature), a third single charge / discharge process at 45°C (fourth temperature), and a fourth single charge / discharge process at 0°C (third temperature). Examples of electrochemical parameters obtained by the first to fourth single charge / discharge processes and cumulative prediction accuracy are shown in Table 2 below.

[0038]

[0039] When the evaluation charge / discharge protocol including the first to fourth single charge / discharge processes is compared with the IEC 62660-1 test as a predetermined charge / discharge protocol, the current value of 1C for evaluation A is three times the current value of 1 / 3C for the energy density step of IEC 62660-1, thereby shortening the discharge time to one-third (1 hour). Furthermore, because the evaluation charge / discharge protocol performs power density evaluations B to D during the process of performing energy density evaluation A, the time required for the power density measurement step of IEC 62660-1 can be significantly reduced. Furthermore, the evaluation charge / discharge protocol is controlled to terminate when the user's desired prediction accuracy is achieved. For example, if the user desires a prediction accuracy of 80% and the prediction accuracy exceeds 80% upon completion of the third single charge / discharge process, the evaluation charge / discharge protocol can be terminated at that point, thereby eliminating the time required for the fourth single charge / discharge process. Even if all of the above-mentioned first to fourth single charge / discharge processes are performed, the evaluation charge / discharge protocol takes less than one day (including the time required for adjusting the storage battery between multiple single charge / discharge processes).

[0040] <Charging / Discharging Device> Figure 3 is a diagram showing an example of the evaluation result prediction system 1 according to the embodiment of the present disclosure shown in Figure 1. In Figure 3, the charging / discharging device 100 includes one or more storage batteries 101 (cell 1 to cell M) connected in series or parallel, a power supply unit 102 electrically connected to the storage batteries, and a sensor unit 103 that measures test condition data and charge / discharge data for each storage battery. The charging / discharging device 100 further includes a temperature adjustment device (not shown) that can adjust the temperature of the environment surrounding the storage batteries being charged / discharged to any desired temperature.

[0041] The storage battery 101 is a rechargeable secondary battery, such as a lithium-ion secondary battery (LiB), a lead-acid battery, or a nickel-metal hydride secondary battery, preferably an LiB. When the charging / discharging device 100 includes multiple storage batteries, each battery can be individually controlled. For example, the device can determine whether to continue testing for each battery based on the content of the obtained charging / discharging data, such as by halting the charging / discharging protocol for only some storage batteries and continuing the charging / discharging protocol for the other storage batteries. After the evaluation, the storage battery 101 may be disassembled and analyzed or instrumentally analyzed to obtain material analysis data at that time. If no material analysis data is obtained, the storage battery 101 may be returned to the user. Alternatively, the storage battery 101 may be used to continue the life evaluation test and perform an evaluation charging / discharging protocol and / or material analysis to obtain additional charging / discharging data and / or material analysis data, or may be used to obtain new training data (evaluation results) by performing a further predetermined charging / discharging protocol. The power supply unit 102 can implement an evaluation charging / discharging protocol on the storage battery. When acquiring training data for creating a database, which will be described later, the power supply unit 102 can also perform a predetermined life evaluation test or a predetermined charge / discharge protocol on the storage battery.

[0042] The sensor unit 103 may include, for example, a voltage sensor, an ammeter, and a temperature sensor. Examples of test condition data measured by the sensor unit 103 include the temperature, current value, voltage (e.g., upper limit voltage, lower limit voltage), charge / discharge pause time, changes over time of the storage battery, and mathematically processed data. Examples of charge / discharge data measured by the sensor unit 103 include the voltage, current, temperature, changes over time of the storage battery, and mathematically processed data. The sensor unit 103 may measure the test condition data and charge / discharge data in any manner. For example, voltage measurements can be performed by attaching voltage sensor terminals to the positive and negative terminals of the target storage battery. Current measurements can be performed by attaching an ammeter to the target wiring. Temperature measurements can be performed by attaching a temperature sensor to the target storage battery. The test condition data and charge / discharge data acquired by the sensor unit 103 are transmitted to the charging / discharging device terminal 200.

[0043] <Charging / Discharging Device Terminal> In FIG. 3 , the charging / discharging device terminal 200 includes an acquisition unit 201 that acquires charging / discharging data of each storage battery measured by the sensor unit 103 of the charging / discharging device 100, a transmission / reception unit 205 that transmits / receives charging / discharging data and charging / discharging control commands, etc. to / from the business operator server 300 and the business operator terminal 500 via the network N1, and a control unit 206 that controls the power supply unit 102 of the charging / discharging device 100 based on the charging / discharging control commands. The functional units, such as the acquisition unit, the transmission / reception unit, and the control unit, are connected to each other via a bus and / or an interface unit. The charging / discharging device terminal 200 may be, for example, an information processing device such as a personal computer, a notebook PC, or a tablet PC that is connected one-to-one to the charging / discharging device 100. While FIGS. 1 and 3 show one pair of the charging / discharging device 100 and the charging / discharging device terminal 200, the evaluation result prediction system 1 may include multiple pairs of the charging / discharging device 100 and the charging / discharging device terminal 200.

[0044] The acquisition unit 201 of the charge / discharge device terminal 200 acquires charge / discharge data from the charge / discharge device of the storage battery. The acquisition unit acquires (receives) the charge / discharge data directly from the charge / discharge device via a wired or wireless connection. Alternatively, the acquisition unit may acquire (receive) the charge / discharge data via the network N1. The acquired charge / discharge data may be stored in a storage unit (not shown) of the charge / discharge device terminal 200.

[0045] The transmitting / receiving unit 205 of the charging / discharging device terminal 200 transmits the charging / discharging data to the acquiring unit (receiving unit) 301 of the business operator server 300 via the network N1. Alternatively, the transmitting / receiving unit may transmit the charging / discharging data to the acquiring unit (receiving unit) 301 of the business operator server 300 via a wired connection or a wireless connection. The transmitting / receiving unit also receives a command to control the charging / discharging device 100 from the business operator server 300 via the network N1 or via a wired connection or a wireless connection.

[0046] The control unit 206 of the charging / discharging device terminal 200 can control the charging / discharging device 100 so that the power supply unit of the charging / discharging device executes a charge / discharge cycle in a predetermined life evaluation test, a predetermined charge / discharge protocol, or an evaluation charge / discharge protocol for a storage battery. The control unit 206 can also control the power supply unit of the charging / discharging device so as to end or continue the evaluation charge / discharge protocol based on a command received from the business operator server 300. The control of the power supply unit is executed based on a computer program (charge / discharge control software) stored in a storage unit (not shown) of the charging / discharging device terminal 200.

[0047] 3 functions as an "evaluation result prediction device" that predicts the evaluation results of a predetermined life evaluation test and / or predicts the energy density and power density of a storage battery that are sequentially measured using a predetermined charge / discharge protocol, and also functions as a server. The business operator server 300 may further function as an operable factor prediction device that predicts operable factors, which are design factors of a storage battery that affect the life of the storage battery. The business operator server 300 includes an acquisition unit (receiving unit) 301 that acquires test condition data, charge / discharge data, material analysis data, etc. of the storage battery, an explanatory variable extraction unit 302 that extracts multiple explanatory variables from the test condition data, charge / discharge data, and material analysis data, a storage unit 303 that stores a trained model (also referred to as a “trained model I” in this disclosure) that has been trained to output a predicted evaluation result when multiple explanatory variables are input, and a database, etc., an evaluation prediction unit 304 that acquires a predicted evaluation result by inputting the multiple explanatory variables to the trained model and calculates the prediction accuracy of the predicted evaluation result, and an output unit (transmitting unit) 305 that outputs information including the predicted evaluation result and the prediction accuracy. The business operator server 300 in FIG. 3 further includes a control unit 306 that outputs a command to (i) terminate the evaluation charge / discharge protocol if the prediction accuracy is equal to or greater than a predetermined threshold, and (ii) continue the evaluation charge / discharge protocol if the prediction accuracy is less than the predetermined threshold. The business operator server 300 in FIG. 3 further includes a learning processing unit 307 that can generate and re-learn a trained model.

[0048] The functional units, such as the acquisition unit (reception unit), explanatory variable extraction unit, memory unit, evaluation prediction unit, output unit (transmission unit), control unit, and learning processing unit, are connected to each other via a bus and / or an interface unit. The explanatory variable extraction unit, evaluation prediction unit, control unit, and learning processing unit are implemented by one or more processing units, such as a CPU (Central Processing Unit), that operate based on a program pre-stored in the memory unit. The processing unit may be a digital signal processor (DSP), large scale integration (LSI), application specific integrated circuit (ASIC), field-programmable gate array (FPGA), or the like. The processing unit is connected to the acquisition unit (reception unit), memory unit, output unit (transmission unit), and the like, and controls the functional units.

[0049] The provider server may be configured with a single server or multiple servers connected to each other so that they can communicate with each other. When the provider server is configured with multiple servers, each functional unit may be implemented in a different server.

[0050] Acquisition unit (reception unit): The acquisition unit (reception unit) 301 of the business operator server 300 has a wired communication interface circuit that complies with a communication protocol such as TCP / IP (Transmission Control Protocol / Internet Protocol), and is communicatively connected to the networks N1 and N2 in accordance with a communication standard such as Ethernet (registered trademark). The acquisition unit (reception unit) may also have an antenna that transmits and receives wireless signals and a wireless communication interface circuit that complies with a communication protocol such as wireless LAN, and be communicatively connected to the networks N1 and N2 in accordance with a communication standard such as wireless LAN.

[0051] The acquisition unit (receiving unit) acquires (receives) the test condition data and the charge / discharge data from the charging / discharging device terminal via the network N1. Alternatively, the acquisition unit (receiving unit) may acquire (receive) the test condition data and the charge / discharge data from the charging / discharging device terminal via a wired or wireless connection. The test condition data are specific test conditions for the life evaluation test, including, for example, the temperature, current value, voltage (e.g., upper limit voltage, lower limit voltage), and charge / discharge rest time of the storage battery. The test condition data may be test condition data as actual measured values ​​measured from the storage battery during the life evaluation test, or may be test condition data as set values. From the perspective of improving prediction accuracy, it is preferable to acquire the test condition data as actual measured values. When acquiring the test condition data as set values, the acquisition unit (receiving unit) may acquire (receive) the test condition data as set values ​​from the business operator terminal 500 via the network N1 or from the customer terminal 600 via the virtual environment and the network N2. Alternatively, the acquisition unit (receiving unit) may acquire test condition data as set values ​​pre-stored in a memory unit (not shown) of the charging / discharging device terminal 200 or the memory unit 303 of the business operator server 300. The charge / discharge data is charge / discharge data measured by implementing a predetermined charge / discharge protocol on the storage battery, and includes, for example, the voltage, current, and temperature of the storage battery, their changes over time, and data obtained by mathematically processing these data. Furthermore, the acquisition unit (receiving unit) acquires (receives) material analysis data of the storage battery from the business operator terminal 500 or the like via the network N1. The material analysis data is data obtained by analyzing the components of the target storage battery (such as the positive electrode, negative electrode, separator, and electrolyte) and the materials constituting these components (such as the positive electrode active material, negative electrode active material, separator material, and electrolyte material). The material analysis data may be in any format, such as an X-ray diffraction (XRD) spectrum, a nuclear magnetic resonance (NMR) spectrum, an electron spin resonance (ESR) spectrum, a scanning electron microscope (SEM) image, an ion chromatography (IC) spectrum, etc. The acquired (received) test condition data, charge / discharge data, and material analysis data are stored in the storage unit 303 of the business operator's server.

[0052] Explanatory variable extraction unit: The explanatory variable extraction unit 302 of the business operator server 300 extracts multiple explanatory variables to be used as inputs for the trained model from the test condition data, charge / discharge data, and material analysis data stored in the storage unit. When predicting the evaluation results of a predetermined life evaluation test, the multiple explanatory variables include one or more test condition parameters selected from the test condition data and electrochemical parameters obtained by electrochemically analyzing the charge / discharge data. The multiple explanatory variables optionally further include one or more mathematical parameters obtained by mathematically processing the shape of the discharge curve in the charge / discharge process. The multiple explanatory variables optionally further include one or more material analysis parameters obtained by mathematically processing material analysis data of the storage battery. When predicting the evaluation results of a predetermined charge / discharge protocol, the multiple explanatory variables include electrochemical parameters obtained by electrochemically analyzing the charge / discharge data, and optionally further include one or more mathematical parameters obtained by mathematically processing the shape of the discharge curve in the charge / discharge process.

[0053] The test condition parameters can be directly used as the test condition parameters without processing the test condition data. That is, the test condition parameters can be selected from the temperature, current value, voltage (e.g., upper limit voltage, lower limit voltage), and charge / discharge rest time of the storage battery in the life evaluation test. However, parameters mathematically processed from the test condition data can also be used as the test condition parameters.

[0054] The electrochemical parameters include energy density and power density, and optionally further include data related to at least one selected from voltage relaxation behavior and capacity retention rate. The electrochemical parameters can be extracted by electrochemically analyzing the charge / discharge process of the evaluation charge / discharge protocol. For example, the energy density can be calculated from the cumulative current capacity (Wh) when the storage battery is discharged from 100% SOC to 0% and the battery weight (kg) or battery volume (L) using the following formula: gravimetric energy density (Wh / kg) = cumulative current capacity (Wh) / battery weight (kg), or the following formula: volumetric energy density (Wh / L) = cumulative current capacity (Wh) / battery volume (L). The power density can be calculated by temporarily calculating the maximum current value I ｍａｘ The voltage drop (V) after discharge when (A) is output, and the battery weight (kg) or battery volume (L) are used to calculate the weight power density (W / kg) = voltage drop (V) × current value I ｍａｘ (A) / battery weight (kg), or the following formula: volume power density (W / L) = voltage drop (V) × current value I ｍａｘ (A) / battery volume (L). The battery weight (kg) or battery volume (L) is acquired separately from the charge / discharge data. For example, this information is measured before connecting the storage battery to the charge / discharge device 100, and the measurement value is input to the charge / discharge device terminal 200 or the business operator terminal 500. Information regarding the battery weight (kg) or battery volume (L) is transmitted from the charge / discharge device terminal 200 or the business operator terminal 500 to the acquisition unit (receiving unit) 301 of the business operator server 300 via the network N1. The business operator server 300 may store the acquired (received) information regarding the battery weight (kg) or battery volume (L) in the storage unit 303. Then, the explanatory variable extraction unit 302 of the business operator server 300 uses the information regarding the battery weight (kg) or battery volume (L) to calculate the energy density and power density as described above. The voltage relaxation behavior is defined as the maximum current value I ｍａｘ The behavior of overvoltage relaxation when charging and discharging are temporarily stopped after the output is parameterized, and is expressed as, for example, the following formula: Voltage relaxation (V) = Voltage (V) after charging and discharging are stopped for 1 minute - Maximum current value I ｍａｘThe capacity retention rate is calculated from the voltage (V) after outputting the voltage (V) at each temperature. For example, when the evaluation charge / discharge protocols shown in Table 2 above are used, the energy density, power densities at SOC 80%, SOC 50%, and SOC 20%, and voltage relaxation behavior are obtained from the first to fourth single charge / discharge processes at each temperature. The capacity retention rate can be calculated from the discharge capacity of the storage battery in a new state (at the start of the life evaluation test) under the conditions for implementing the life evaluation test (e.g., the integrated current capacity when discharging from SOC 100% to 0%) and the discharge capacity of the storage battery at a measurement point during the life evaluation test (e.g., the point at which a single charge / discharge process of the evaluation charge / discharge protocol was implemented) using the following formula: capacity retention rate = (discharge capacity of the storage battery at the measurement point during the life evaluation test / discharge capacity of the storage battery at the start of the life evaluation test) × 100. For example, when an evaluation charge / discharge protocol including performing a single charge / discharge process illustrated in FIG. 2( a) for a certain period or for a certain number of charge / discharge cycles is used, the energy density, the power density at SOC 80%, SOC 50%, and SOC 20%, the voltage relaxation behavior, and the capacity retention rate can be obtained from each single charge / discharge process.

[0055] The mathematical parameters can be extracted by mathematically processing the shape of the discharge curve in the charge / discharge process. The mathematical parameters include, for example, data obtained by mathematically processing a discharge curve in which the vertical axis represents the voltage of the storage battery and the horizontal axis represents any combination of time, current capacity, SOC, etc. More specifically, as the mathematical parameters, a matrix X with 1 row and B columns is first obtained, in which voltage (V) values ​​corresponding to B (B≧2) different SOC values ​​are arranged as elements in the column direction for discharge curve data in which the SOC value of the storage battery to be evaluated is represented on the horizontal axis and the corresponding voltage (V) value is represented on the vertical axis. The number B of SOC values ​​used is approximately 10 to 1000, corresponding to the number B of SOC values ​​used in generating the trained model. For example, as illustrated in FIG. 4 , if the SOC value is divided in 1% increments from 100% to 0%, B is a total of 101, including 100%, 99%, 98%, ..., 2%, 1%, and 0%. The number of rows in matrix X is 1 because the rating prediction process is performed for each storage battery (for each storage battery). Next, after applying mathematical constraints, matrix X is decomposed into X≈W×H (where matrix W is a coefficient matrix with 1 row and C columns, and matrix H is a basis matrix with C rows and B columns). The mathematical constraints are set so that matrices W and H are uniquely determined. C corresponds to C used when generating the trained model and is, for example, equal to or less than the number B of SOC values. Methods for matrix decomposition include principal component analysis and nonnegative matrix factorization. Then, one or more explanatory variables can be selected from the C coefficients in matrix W (referred to as the "first principal component" to the "Cth principal component," starting from the first column).

[0056] The material analysis parameters can be obtained by mathematically processing material analysis data of the storage battery, such as an X-ray diffraction (XRD) spectrum, a nuclear magnetic resonance (NMR) spectrum, an electron spin resonance (ESR) spectrum, a scanning electron microscope (SEM) image, or an ion chromatography (IC) spectrum of the material that constitutes the storage battery. Methods for obtaining the material analysis parameters from the material analysis data include the following processing steps (1) to (3).

[0057] Processing process (1) is a method for acquiring material analysis parameters from spectra such as XRD, NMR, ESR, and IC. First, a matrix Y is obtained with 1 row and E columns (E is the number of elements, where E is ≥ 2) in which the spectral data in a new state and the spectral data at the time of analysis are arranged as elements in the column direction. The difference between the spectral data in a new state and the spectral data at the time of analysis may be used. The number of elements E employed is approximately 10 to 1000, corresponding to the number of divisions E in the horizontal direction of the spectral data before matrix decomposition. The number of rows in matrix Y is 1 because the evaluation prediction process is performed for each storage battery (for each storage battery). Next, after applying mathematical constraints, matrix Y is decomposed into Y≈V×I (where matrix V is a coefficient matrix with 1 row and F columns, and matrix I is a base matrix with F rows and E columns). The mathematical constraints are set so that matrices V and I are uniquely determined. F is, for example, equal to or less than the number of elements E. As a matrix decomposition method, principal component analysis, non-negative matrix factorization, etc. can be used. For example, when principal component analysis is performed, F coefficients in matrix V (referred to as the "first principal component" to the "Cth principal component" in order from the first column) can be obtained as material analysis parameters. Then, one or more explanatory variables can be selected from the obtained material analysis parameters.

[0058] Processing step (2) is another method for acquiring material analysis parameters from spectra such as XRD, NMR, ESR, and IC. First, parameters characteristic of the curve outline, such as the intensity, integral value, and half-width of the peaks that characterize the material being analyzed, are calculated from the spectral data in the new state and at the time of analysis. Then, the difference between the parameters in the new state and the parameters at the time of analysis is calculated, and the obtained difference can be acquired as material analysis parameters. Then, one or more explanatory variables can be selected from the acquired material analysis parameters.

[0059] Processing (3) is a method of acquiring material analysis parameters from image data such as SEM images. First, a filter process such as a Gaussian filter is applied to the image data in the new state and at the time of analysis to remove noise. Next, a binarization process such as Otsu's binarization is applied to calculate characteristic parameters such as the area ratio of bright and dark areas and circularity. Then, for example, the difference between the parameters in the new state and the parameters at the time of analysis is calculated, and the obtained difference can be acquired as material analysis parameters. Then, one or more explanatory variables can be selected from the acquired material analysis parameters.

[0060] The explanatory variables may be pre-selected to improve the prediction accuracy of the trained model and / or improve the interpretability of the prediction evaluation results. More specifically, the explanatory variables may be pre-selected by excluding one or more parameters that are multicollinear with other parameters when generating the trained model.

[0061] For example, when the evaluation charge / discharge protocol shown in Table 2 above is used and B = C = 101 as illustrated in FIG. 4 , a total of 106 parameters can be extracted from one single charge / discharge process for the storage battery being evaluated, including five electrochemical parameters (energy density, power density at 80% SOC, power density at 50% SOC, power density at 20% SOC, and voltage relaxation behavior) and 101 mathematical parameters (first principal component to first principal component). When all of the first to fourth single charge / discharge processes are performed, a total of 424 parameters (4 temperatures x 106) are extracted. While all of these parameters may be used as input variables for the trained model, including multiple parameters that are strongly correlated with each other, multicollinearity can actually reduce prediction accuracy and / or interpretability. Therefore, prediction accuracy and / or interpretability can be improved by using pre-selected explanatory variables as input variables, in which one or more parameters that are multicollinear with other parameters are manually or using some variable selection method to eliminate them. For example, examples of pre-selected explanatory variables when the evaluation charge / discharge protocol shown in Table 2 above is used are shown in Table 3 below. The charge / discharge data obtained by implementing the evaluation charge / discharge protocol was subjected to matrix decomposition using principal component analysis as described below, and each of the obtained principal components was used as a candidate explanatory variable. The combinations of explanatory variables and prediction accuracy shown in Table 3 below are merely examples.

[0062]

[0063] As shown in Table 3 above, by using pre-selected explanatory variables, the user's desired prediction accuracy can be achieved with fewer charge / discharge processes, further shortening the time required for the evaluation prediction process. For example, if the user desires a prediction accuracy of 80%, the evaluation charge / discharge protocol can be terminated upon completion of the third single charge / discharge process, thereby eliminating the time required for the fourth single charge / discharge process. Furthermore, by using pre-selected explanatory variables, it becomes easier to understand the factors contributing to the objective variable, improving the interpretability of the calculation results obtained by machine learning.

[0064] Examples of combinations of explanatory variables include combinations of one or more test condition parameters and one or more electrochemical parameters; one or more test condition parameters, one or more electrochemical parameters, and one or more mathematical parameters; one or more test condition parameters, one or more electrochemical parameters, and one or more material analysis parameters; and one or more test condition parameters, one or more electrochemical parameters, one or more mathematical parameters, and one or more material analysis parameters. In this way, by using a multimodal trained model that has learned data of different properties obtained from multiple information sources, highly interpretable (white-box) predictive evaluation results based on the principles of battery degradation can be obtained. Furthermore, by using a trained model that simultaneously trains multiple related explanatory variables (multitask learning), the predictive accuracy of the predictive evaluation results can be further improved. Among the above combinations of explanatory variables, a combination including test condition parameters, electrochemical parameters, and material analysis parameters is preferred because it allows for obtaining predictive results that combine predictive accuracy and interpretability even with limited data.

[0065] Storage unit: The storage unit 303 of the business operator server 300 includes a memory device such as a RAM (Random Access Memory) or a ROM (Read Only Memory), a fixed disk device such as a hard disk, or a portable storage device such as a flexible disk or an optical disk. The storage unit can store computer programs used for various processes of the business operator server 300, acquired charge / discharge data, trained models, databases used for training processes of the trained models, and the like. The computer programs may be installed into the storage unit from a computer-readable portable recording medium using a known setup program or the like. Examples of portable recording media include a CD-ROM (compact disc read only memory) and a DVD-ROM (digital versatile disc read only memory). The computer programs may be stored in a recording medium owned by a specific server and installed via the networks N1 and / or N2.

[0066] When predicting the evaluation results of a predetermined life evaluation test, the trained model is trained to output a predicted evaluation result including an evaluation result that would be obtained if the predetermined life evaluation test were performed and predicted deterioration information of the storage battery at the end of its life when multiple explanatory variables extracted from the charge / discharge data of the storage battery are input.When predicting the evaluation results of a predetermined charge / discharge protocol, the trained model is trained to output a predicted evaluation result including the energy density and power density measured in the predetermined charge / discharge protocol when multiple explanatory variables extracted from the charge / discharge data of the storage battery are input.The trained model may be trained to predict the evaluation results of the predetermined life evaluation test and to predict the evaluation results of the predetermined charge / discharge protocol.The multiple explanatory variables may be pre-selected explanatory variables as described above.

[0067] The database stores a dataset used to generate the trained model. When predicting the evaluation results of a predetermined life evaluation test, the dataset includes evaluation results as training data obtained by conducting the predetermined life evaluation test on multiple storage batteries and deterioration information of the storage batteries as training data obtained by analyzing the materials of the storage batteries at the end of their life. The dataset further includes test condition data for the life evaluation test, charge / discharge data obtained by implementing an evaluation charge / discharge protocol on multiple storage batteries in various states of deterioration, and material analysis data for the storage batteries obtained by dismantling and analyzing the multiple storage batteries in various states of deterioration. The database also includes multiple explanatory variables extracted from the test condition data, charge / discharge data, and material analysis data, which are linked to the training data and used to train the machine learning model. When predicting the evaluation results of a predetermined charge / discharge protocol, the dataset includes evaluation results, including energy density and power density, as training data obtained by implementing the predetermined charge / discharge protocol on multiple storage batteries, and multiple explanatory variables extracted from the charge / discharge data obtained by implementing the evaluation charge / discharge protocol on the multiple storage batteries and linked to the evaluation results. When predicting the evaluation results of a predetermined life evaluation test and the evaluation results of a predetermined charge / discharge protocol, the data set stores the data set used for learning both. The database is updated as needed and used to retrain the trained model. Details of database construction will be described later.

[0068] The storage unit 303 of the business operator server 300 stores (accumulates) test condition data obtained when a life evaluation test is conducted on the storage battery under evaluation, charge / discharge data obtained by executing the evaluation charge / discharge protocol, and material analysis data obtained by dismantling the storage battery under evaluation. Furthermore, for the storage battery under evaluation, if a life evaluation test is continued after the evaluation prediction is completed to perform the evaluation charge / discharge protocol and / or material analysis, and additional charge / discharge data and / or material analysis data is obtained, the charge / discharge data and / or material analysis data is stored (added) to the database. If a predetermined life evaluation test is conducted on a storage battery from the same lot as the storage battery under evaluation and evaluation results and / or deterioration information are obtained as new training data, the new training data is linked to the extracted explanatory variables and stored (added) to the database. Furthermore, if a predetermined charge / discharge protocol is further conducted on the storage battery under evaluation and evaluation results are obtained as new training data, the new training data is linked to the explanatory variables extracted from the charge / discharge data and stored (added) to the database.

[0069] Evaluation prediction unit: The evaluation prediction unit 304 of the business operator server 300 obtains a predicted evaluation result by inputting multiple explanatory variables extracted from the test condition data, charge / discharge data, and material analysis data into a trained model, and calculates the prediction accuracy of the predicted evaluation result.

[0070] The predicted evaluation results include evaluation results that would be obtained if a predetermined life evaluation test were performed on the storage battery being evaluated, and predicted deterioration information of the storage battery at the end of its life, and / or include energy density and power density that would be obtained if a predetermined charge / discharge protocol were performed on the storage battery being evaluated. Examples of evaluation results that would be obtained if a predetermined life evaluation test were performed include the storage battery's post-test voltage retention rate, post-test capacity retention rate, life time, and number of cycles to life. The life time may be, for example, the time required for the storage battery's capacity retention rate to reach 80%. The number of cycles to life may be, for example, the number of cycles required for the storage battery's capacity retention rate to reach 80%. Furthermore, the predicted deterioration information of the storage battery includes the degree of deterioration of the storage battery's materials and components at the end of its life (e.g., when the storage battery's capacity retention rate reaches 80% or when the storage battery's temperature reaches an upper limit). The degree of material deterioration is expressed as a percentage of the remaining functioning health of each material at the end of the battery life, with the new state (at the start of the life evaluation test) being 100% for the materials of the battery, such as the positive electrode active material, the negative electrode active material, the separator material, and the electrolyte material (solvent, salt, and additives). The degree of component deterioration indicates the health of the battery components, such as the positive electrode, the negative electrode, the separator, and the electrolyte, at the end of the battery life. Examples of component deterioration include the degree of pore blockage (the rate of reduction in the number of pores) of the separator expressed as a percentage with the new state being 0%, a binary evaluation of whether the separator is broken or not, and the rate of decrease in separator strength expressed as a percentage with the new state being 100%. In this way, the predicted evaluation result includes predicted deterioration information of the storage battery at the end of its life, which has the advantage that the user can know the cause of deterioration of the storage battery, consider the deterioration mechanism, and plan countermeasures against deterioration, etc. If there are multiple measurement conditions for the energy density and power density to be predicted, the energy density and power density that would be obtained if a predetermined charge / discharge protocol is performed include at least one of them, and preferably all of them.For example, when IEC 62660-1 is used as the predetermined charge / discharge protocol to be predicted, the predicted evaluation results can include (1) energy density at 25°C, (2) power density at 25°C and SOC 20%, (3) power density at 25°C and SOC 50%, (4) power density at 25°C and SOC 80%, (5) power density at -20°C and SOC 50%, (6) power density at 0°C and SOC 50%, and (7) power density at 45°C and SOC 50%. The predicted evaluation results include at least one predicted evaluation result selected from the group consisting of (1) and (2) to (7). The predicted evaluation results preferably include the predicted evaluation results of (1) to (4), and more preferably include all of the predicted evaluation results of (1) to (7).

[0071] The prediction accuracy of the evaluation results of a predetermined life evaluation test indicates the degree of accuracy of the predicted evaluation results with respect to the evaluation results and deterioration information when the predetermined life evaluation test is performed on the storage battery to be evaluated and material analysis is performed at the end of the life. The coefficient of determination (R ２ The prediction accuracy can be calculated by using the following methods: (1) a linear regression function (LDF), (2) a linear regression function (LDF), (3) a linear regression function (LDF), (4) a linear regression function (LDF), (5) a linear regression function (LDF), (6) a linear regression function (LDF), (7) a linear regression function (LDF), (8) a linear regression function (LDF), (9) a linear regression function (LDF), (10) a linear regression function (LDF), (11) a linear regression function (LDF), (12) a linear regression function (LDF), (13) a linear regression function (LDF), (14) a linear regression function (LDF), (15) a linear regression function (LDF), (16) a linear regression function (LDF), (17) a linear regression function (LDF), (18) a linear regression function (LDF), (19) a linear regression function (LDF), (20) a linear regression function (LDF), (21) a linear regression function (LDF), (22) a linear regression function (LDF), (23) a linear regression function (LDF), (24) a linear regression function (LDF), (25) a linear regression function (LDF), (26) a linear regression function (LDF), (27) a linear regression function (LDF), (28) a linear regression function (LDF), (29 ... ２The prediction accuracy can be calculated using various methods, such as the root mean square error (RMSE), mean absolute error (MAE), and mean absolute percentage error (MAPE). The prediction accuracy varies depending on which charging / discharging processes in the evaluation charging / discharging protocol have been completed (the degree of accumulation of charging / discharging data) and the combination of input explanatory variables. The more charging / discharging processes are completed and the more charging / discharging data is accumulated, the higher the prediction accuracy. In addition, the prediction accuracy can be improved by eliminating one or more explanatory variables that are multicollinear with other explanatory variables (Tables 2 and 3).

[0072] Output unit (transmitter): The output unit (transmitter) 305 of the business operator server 300 has a wired communication interface circuit that complies with a communication protocol such as TCP / IP (Transmission Control Protocol / Internet Protocol), and is communicatively connected to the networks N1 and N2 in accordance with a communication standard such as Ethernet (registered trademark). The acquisition unit (receiver) 301 may have an antenna that transmits and receives wireless signals and a wireless communication interface circuit that complies with a communication protocol such as wireless LAN, and may be communicatively connected to the networks N1 and N2 in accordance with a communication standard such as wireless LAN.

[0073] The output unit (transmission unit) 305 of the business operator server 300 outputs (transmits) information including the predicted evaluation results and prediction accuracy (hereinafter simply referred to as "prediction information") to the acquisition unit 401 of the virtual environment 400 accessible by the user terminal via the network N2. The output (transmission) of the prediction information is performed in real time. This allows the user to check the storage battery prediction information in real time and control the evaluation result prediction device. Note that in this disclosure, "outputting" the prediction information includes "transmitting" the prediction information to another device and "displaying" the prediction information in any form, such as a graphical user interface (GUI), on a display unit (not shown) of the business operator server 300 or another device. In the business operator server 300 of FIG. 3, "outputting" the prediction information means "transmitting" it to the virtual environment 400. The output unit (transmission unit) 305 may output (transmit) the prediction information to the business operator terminal 500 and / or the customer terminal 600. The business entity server 300 may further have a functional unit for "displaying" on a display unit (not shown) of its own or another device. The display unit may be a display including a liquid crystal display, an organic electroluminescence (EL) display, or the like.

[0074] Control Unit: The control unit 306 of the business operator server 300 can control the charge / discharge device 100 via the charge / discharge device terminal 200 to terminate or continue the evaluation charge / discharge protocol (terminate or continue the life evaluation test). For example, after performing a certain charge / discharge process, the control unit 306 can (i) terminate the evaluation charge / discharge protocol (terminate the life evaluation test) if the prediction accuracy is equal to or greater than a predetermined threshold, or (ii) continue the evaluation charge / discharge protocol (continue the life evaluation test) if the prediction accuracy is less than the threshold, and output a command to perform the next charge / discharge process. In the case of (ii), the acquisition unit acquires further charge / discharge data obtained from the next charge / discharge process, and optionally further material analysis data, and the explanatory variable extraction unit extracts explanatory variables from the accumulated charge / discharge data and material analysis data as described above. The evaluation prediction unit inputs the explanatory variables extracted from the accumulated charge / discharge data and material analysis data into the trained model, and obtains predicted evaluation results and calculates prediction accuracy as described above. The control unit 306 then compares the prediction accuracy of the predicted evaluation results obtained based on the accumulated charge / discharge data and material analysis data with a predetermined threshold, and (i) if the prediction accuracy is equal to or greater than the predetermined threshold, terminates the evaluation charge / discharge protocol (terminates the life evaluation test), or (ii) if the prediction accuracy is less than the threshold, continues the evaluation charge / discharge protocol (continues the life evaluation test) and outputs a command to perform the next charge / discharge process. The control unit 306 can repeat the above process until the prediction accuracy is equal to or greater than the predetermined threshold or until the nth (all) charge / discharge process is completed. In this way, the business operator server 300 can automatically control the termination or continuation of the evaluation charge / discharge protocol based on the prediction accuracy, and can more quickly feed back predicted evaluation results that meet the prediction accuracy required by the user.

[0075] The business operator server 300 can perform automatic control using the control unit 306 described above, and may be further configured to control the evaluation result prediction device when it receives an instruction (an instruction from a user) to control the evaluation result prediction device from the virtual environment 400. Examples of instructions from the user that control the evaluation result prediction device include an instruction to set or change a threshold value for prediction accuracy, an instruction to end or continue the evaluation charge / discharge protocol (an instruction to end or continue the life evaluation test), etc. Control of the evaluation result prediction device using the virtual environment 400 will be described in detail below.

[0076] Learning Processing Unit: The learning processing unit 307 of the business operator server 300 performs learning processing of a machine learning model based on a machine learning program, and can generate and re-learn a trained model. Details of the learning processing of a machine learning model will be described later.

[0077] 3 , the virtual environment 400 is a virtual server constructed on the network N2 and is communicatively connected to the provider server 300 via the network N2. The virtual environment 400 may be a virtual server constructed using the hardware resources of the provider server 300, or may be an environment constructed using the hardware resources of one or more servers other than the provider server 300. The virtual environment 400 includes an acquisition unit 401 that acquires information including a predicted evaluation result and prediction accuracy, an output unit 405 that outputs (displays) the information including the predicted evaluation result and prediction accuracy in the form of a graphical user interface (GUI), and a control unit 406 that outputs a command to control the evaluation result prediction device when a command to control the evaluation result prediction device is input from a user.

[0078] The acquisition unit 401 of the virtual environment 400 can acquire information including the predicted evaluation results and prediction accuracy (hereinafter simply referred to as "prediction information"), and optionally information including charge / discharge data, etc., from the business operator server 300 via the network N2 in real time. The output unit (GUI) 405 of the virtual environment 400 can output (display) the prediction information acquired by the acquisition unit 401 in real time in the form of a GUI on the display device (display unit) of the terminal device (business operator terminal 500, customer terminal 700, etc.) that accesses the virtual environment 400. Information displayed on the GUI includes, for example, the storage battery's SOC, voltage and current values, the progress of a life evaluation test, the current predicted evaluation results and prediction accuracy, and the prediction accuracy for the next charge / discharge process. The GUI can display this information graphically.

[0079] The GUI not only outputs (displays) prediction information, etc., but also functions as an input unit for inputting commands from the user to control the evaluation result prediction device. Examples of commands include an instruction to set or change a prediction accuracy threshold, an instruction to terminate or continue the evaluation charge / discharge protocol (terminate or continue the life evaluation test), etc. The GUI displays icons, tabs, input bars, etc. corresponding to these commands, allowing the user to input commands via the GUI. When a user inputs a command via the GUI, the control unit 406 of the virtual environment 400 outputs (transmits) a command to control the evaluation result prediction device to the business operator server 300 via the network N2. When multiple storage batteries are being evaluated in parallel, the virtual environment 400 is configured to output (display) prediction information for each storage battery in the form of a GUI and control charging and discharging for each storage battery. This allows the user to check prediction information for multiple storage batteries to be evaluated in real time for each storage battery and control the evaluation result prediction device. For example, the user can change the threshold to a higher value before the prediction accuracy of a certain storage battery reaches a predetermined threshold and continue the evaluation charge / discharge protocol (continue the life evaluation test), or can terminate the evaluation charge / discharge protocol (terminate the life evaluation test) even if the prediction accuracy of the evaluation charge / discharge protocol of a certain storage battery is below the predetermined threshold.

[0080] 3 is one or more information processing devices operated by a battery evaluation contractor, and may be, for example, a personal computer, a notebook PC, a tablet PC, or a multi-function mobile phone (a so-called smartphone). The business operator terminal 500 is communicatively connected to the charging / discharging device terminal 200 and the business operator server 300 via the business operator's private network N1, and can manage, operate, and modify the evaluation result prediction system 1. The business operator terminal 500 is also communicatively connected to the virtual environment 400 and the customer terminal 600 via a user-accessible network N2, and can thereby manage, operate, and modify the virtual environment 400 and provide technical support to customers.

[0081] <Customer Terminal> The customer terminal 600 in FIG. 3 is one or more information processing devices operated by a business entity that has commissioned a battery evaluation. For example, the customer terminal 600 may be a personal computer, a notebook PC, a tablet PC, or a multi-function mobile phone (a so-called smartphone). The customer terminal 600 is communicatively connected to the virtual environment 400 via the network N2 and can display the GUI of the virtual environment 400. Through the GUI of the virtual environment 400, a user can obtain prediction information in real time and can send commands to the virtual environment 400 to control the evaluation result prediction device at any time. The customer terminal 600 is also communicatively connected to the business entity terminal 500 via the network N2 and can receive technical support from the business entity.

[0082] 5 and 6 are flowcharts showing an example of the operation of the learning process of the trained model that outputs the evaluation result that would be obtained if a predetermined life evaluation test were conducted and the predicted evaluation result including predicted deterioration information of the storage battery at the end of its life. Note that the flow of the operation described below is executed based on a program stored in the storage unit 303 of the business operator server 300, mainly by the explanatory variable extraction unit 302 and the learning processing unit 307 included in the processing unit of the business operator server 300, in cooperation with each element.

[0083] 5 shows an example of the operation of a learning process for a trained model that does not involve the collection of material analysis data. First, one or more storage batteries 101 for collecting teacher data (storage batteries for collecting teacher data) are connected in series or parallel to the power supply unit 102 of the charging / discharging device 100, and a predetermined life evaluation test is performed (S101). The storage batteries for collecting teacher data are preferably of the same category (e.g., LiB, lead-acid battery, or nickel-metal hydride secondary battery) as the storage battery to be evaluated, and more preferably, a storage battery that uses at least one of the materials (e.g., positive electrode active material, negative electrode active material, separator material, and electrolyte material) used in the components (e.g., positive electrode, negative electrode, separator, and electrolyte).

[0084] In FIG. 5 , next, evaluation results from a predetermined life evaluation test are obtained (S102). The evaluation results include at least one evaluation result selected from the storage battery's post-test voltage retention rate, post-test capacity retention rate, life time, number of cycles to life, or degradation state at the end of life. The obtained evaluation results are used as training data. When there are multiple evaluation results to be obtained in the predetermined life evaluation test, all of them are basically obtained. For example, when a storage battery storage test according to IEC 62660-1 is performed as the predetermined life evaluation test, the evaluation results include the voltage retention rate after storing the storage battery at 45°C for 28 days, and the capacity retention rates after storing the storage battery at 45°C for 42 days, 84 days, and 126 days. When a cycle test of a storage battery according to IEC 62660-1 is performed, the evaluation results include (i) the temperature and capacity retention rate after each deterioration step, and (ii) when the capacity retention rate of the storage battery reaches 80% or (iii) when the temperature of the storage battery reaches the upper limit temperature, the time and number of cycles (time to reach life and number of cycles to reach life) required until that point. However, for some of the storage batteries used for collecting training data, only some of these evaluation results may be obtained as training data.

[0085] In FIG. 5 , one or more brand-new storage batteries (storage batteries for acquiring explanatory variables) from the same lot (manufactured at the same time using the same materials and under the same conditions) as the storage batteries for collecting teacher data are connected in series or parallel to the power supply unit 102 of the charging / discharging device 100, and a life evaluation test is initiated, and test condition data acquisition begins (S103). The test condition data is acquired (received) by the acquisition unit (receiving unit) 301 of the business operator server 300 via the network N1 from the charging / discharging device terminal 200. The conditions of the life evaluation test are the same as those of the predetermined life evaluation test conducted on the storage batteries for collecting teacher data, except that the evaluation charge / discharge process is performed at the start of the life evaluation test and every certain period or certain number of cycles. The test condition data is used to extract explanatory variables that will later be used as input for the learning process. Therefore, the acquired test condition data basically includes all of the temperature, current value, voltage, and charge / discharge rest time of the storage batteries. However, only a portion of this test condition data may be acquired for some storage batteries for acquiring explanatory variables.

[0086] In FIG. 5 , next, based on the evaluation charge / discharge protocol, an evaluation charge / discharge process is performed at the start of the life evaluation test and at regular intervals or after a certain number of charge / discharge cycles, and charge / discharge data is acquired (S104). The charge / discharge data is measured by the sensor unit 103 of the charge / discharge device 100 and transmitted to the charge / discharge device terminal 200. Step S104 of acquiring charge / discharge data is repeated until charge / discharge data is acquired at the end of the life (e.g., when the capacity retention rate of the storage battery reaches 80% or the temperature of the storage battery reaches an upper limit). Then, a plurality of explanatory variables are calculated (extracted) from the test condition data and the charge / discharge data (S105). The explanatory variables are calculated by the explanatory variable extraction unit 302 of the business operator server 300. The explanatory variables include test condition parameters selected from the test condition data, electrochemical parameters obtained by electrochemically analyzing the charge / discharge data, and mathematical parameters obtained by mathematically processing the shape of the discharge curve in the charge / discharge process.

[0087] The test condition parameters include the temperature, current value, voltage (e.g., upper limit voltage, lower limit voltage), and charge / discharge rest time of the storage battery during the life evaluation test, as well as parameters mathematically processed from these, etc. The obtained test condition parameters each become candidate explanatory variables when determining the "selected explanatory variables."

[0088] The electrochemical parameters can be calculated (extracted) by electrochemically analyzing the charge / discharge process of the evaluation charge / discharge protocol, and include energy density, power density, voltage relaxation behavior, and capacity retention. Details of the calculation of the electrochemical parameters are as described above in the "Explanatory Variable Extraction Section." For example, as illustrated in FIG. 2( b), when an evaluation charge / discharge protocol is used that includes a single charge / discharge process at the start of a life evaluation test and at a certain period or every certain number of charge / discharge cycles during the life evaluation test, the energy density, power density at SOC 80%, SOC 50%, and SOC 20%, voltage relaxation behavior, and capacity retention, where the capacity retention at the start of the life evaluation test is set to 100%, are obtained from each single charge / discharge process. The obtained electrochemical parameters each serve as candidate explanatory variables when determining the "selected explanatory variables."

[0089] Mathematical parameters can be extracted by mathematically processing the shape of a discharge curve in a charge / discharge process. The mathematical parameters include, for example, data obtained by mathematically processing a discharge curve in which the vertical axis represents the voltage of a storage battery and the horizontal axis represents any combination of time, current capacity, SOC, etc. FIG. 4 is a diagram illustrating an example of calculation of mathematical parameters as explanatory variables. As illustrated in FIG. 4 , the mathematical parameters are obtained by matrix decomposition of a matrix representing the general shape of a discharge curve. First, for discharge curve data in which the horizontal axis represents the SOC value of a storage battery and the vertical axis represents the corresponding voltage (V) value, a matrix X with A rows and B columns is obtained in which the voltage (V) values ​​of a plurality of storage batteries (the number of samples is A (A≧2)) corresponding to B (B≧2) different SOC values ​​are arranged as elements in the column direction. The number B of SOC values ​​used may be approximately 10 to 1,000. For example, as illustrated in FIG. 4 , when the SOC value is divided from 100% to 0% in 1% increments, B is 100%, 99%, 98%, ..., 2%, 1%, and 0%, for a total of 101 values. The number of battery samples A is, for example, 10 or more, 50 or more, 100 or more, 500 or more, or 1000 or more. The larger the number of samples A, the more the prediction accuracy of the trained model improves. Next, after applying mathematical constraints, matrix X is decomposed into X≈W×H (where matrix W is a coefficient matrix with A rows and C columns, and matrix H is a basis matrix with C rows and B columns). The mathematical constraints are set so that matrices W and H are uniquely determined. C is, for example, equal to or less than the number B of SOC values. Methods of matrix decomposition that can be used include principal component analysis and nonnegative matrix factorization. The C coefficient matrices in the matrix W (referred to as the "first principal component" to the "Cth principal component" in order from the first column) are each candidates for explanatory variables when determining the "selected explanatory variables."

[0090] Note that the order of conducting the predetermined life evaluation test and obtaining the evaluation results (S101 and S102) and the order of conducting the life evaluation test to obtain the explanatory variables and calculating the explanatory variables (S103 to S105) may be reversed (S103 to S105 first, followed by S101 and S102). Also, the predetermined life evaluation test and the life evaluation test to obtain the explanatory variables may be performed in parallel (S101 and S102, and S103 to S105 in parallel).

[0091] 5, a database including explanatory variables and evaluation results is then created (S106). The database includes a plurality of explanatory variables calculated from the test condition data and charge / discharge data described above, and evaluation results of predetermined life evaluation tests linked to these explanatory variables, and is stored in the storage unit 303 of the business operator's server 300.

[0092] <Generation of Trained Model for Lifespan Assessment Data> Finally, in FIG. 5 , a trained model is generated and stored using the explanatory variables and the assessment results as training data (S107), and the series of processes is terminated. The trained model can be generated by pre-training a machine learning model using supervised learning such as ridge regression, support vector regression, or neural network. As input variables, data obtained by standardizing the above-mentioned explanatory variables (processing to set the average of the data to 0 and the standard deviation to 1) is used. Note that the assessment results as output variables may also be standardized. Next, the machine learning model is trained on the input variables and the corresponding output variables (assessment results), and the prediction accuracy of the model is calculated. As an index of prediction accuracy, the coefficient of determination (R ２ The following can be used: the root mean square error (RMSE), the mean absolute error (MAE), the mean absolute percentage error (MAPE), etc. When verifying the prediction accuracy, it is preferable to perform cross-validation. Then, the hyperparameters are adjusted so that the prediction accuracy of the machine learning model is maximized.

[0093] The input variables are preferably a combination of explanatory variables selected from the plurality of explanatory variables so as to improve the prediction accuracy of the trained model and / or the interpretability of the prediction evaluation results. More specifically, the input variables may be explanatory variables selected by eliminating one or more parameters that are multicollinear with other parameters, rather than using all of the parameters obtained in the calculation of the explanatory variables (S105). Methods for selecting explanatory variables include, for example, methods using Boruta, recursive feature reduction, genetic algorithms (GA), etc. Boruta is a variable selection method based on the variable importance of random forests (RF). The method selects explanatory variables by adding dummy variables to the explanatory variables, constructing a random forest model, calculating the importance of each variable, and removing variables less important than the dummy variables. Recursive feature reduction is a method in which a predictive model is created using all explanatory variables, the least important explanatory variable is removed, a new predictive model is created using the remaining explanatory variables, and the least important explanatory variable is removed again, until a pre-specified number of explanatory variables is reached. Genetic algorithms (GAs) are a method for solving optimization problems, which involves determining the combination of explanatory variables that maximizes accuracy by considering whether each explanatory variable should be used (1 or 0). From the perspectives of predictive accuracy and interpretability, it is preferable to try all of these methods and select the one that provides the best predictive accuracy and interpretability. Examples of combinations of explanatory variables include combinations of one or more test condition parameters and one or more electrochemical parameters; one or more test condition parameters, one or more electrochemical parameters, and one or more mathematical parameters.

[0094] In this manner, a trained model can be generated that is trained to output a predicted evaluation result measured in a predetermined life evaluation test when multiple explanatory variables extracted from the charge / discharge data of the storage battery are input. The output of the generated trained model in FIG. 5 includes, as a predicted evaluation result, at least one selected from the post-test voltage retention rate of the storage battery, the post-test capacity retention rate, the life time, or the number of cycles to life. The generated trained model is stored in the storage unit 303 of the business operator server 300.

[0095] <Creating a Database of Life Evaluation Data> Figure 6 shows an example of the operation of the learning process for a trained model that involves collecting material analysis data. First, one or more storage batteries 101 for collecting teacher data (storage batteries for collecting teacher data) are connected in series or parallel to the power supply unit 102 of the charging / discharging device 100, and a predetermined life evaluation test is performed (S101). Details of step S101 are the same as step S101 in Figure 5 above. Next, evaluation results from the predetermined life evaluation test are obtained (S102). Details of step S102 are the same as step S102 in Figure 5 above.

[0096] 6 , next, a plurality (N) of brand-new storage batteries (explanatory variable acquisition storage batteries) from the same lot (manufactured at the same time using the same materials and under the same conditions) as the teacher data collection storage battery are connected in series or parallel to the power supply unit 102 of the charging / discharging device 100, and a life evaluation test is initiated, and acquisition of test condition data begins (S103). The test condition data is acquired (received) by an acquisition unit (receiving unit) 301 of the business operator server 300 via a network N1 from the charging / discharging device terminal 200. As described below, from the start of the life evaluation test until the end of the life, at least one explanatory variable acquisition storage battery is disassembled each time an evaluation charge / discharge process is performed, and material analysis data and degradation information in various degradation states are acquired. Therefore, the number N of explanatory variable acquisition storage batteries to be prepared is set to be equal to or greater than the number of evaluation charge / discharge processes performed in the life evaluation test. The conditions for this life evaluation test are the same as those for the specified life evaluation test conducted on the storage batteries used to collect training data, except that the evaluation charge / discharge process is performed at the start of the life evaluation test and at regular intervals or after a certain number of cycles. The test condition data is used to extract explanatory variables that will later be used as input for the learning process. Therefore, the acquired test condition data basically includes all of the temperature, current value, voltage, and charge / discharge rest time of the storage battery. However, only a portion of this test condition data may be acquired for some of the storage batteries used to acquire explanatory variables.

[0097] 6 , next, based on the evaluation charge / discharge protocol, an evaluation charge / discharge process is performed at the start of a life evaluation test and at regular intervals or after a certain number of charge / discharge cycles, and charge / discharge data is acquired (S104). The charge / discharge data is measured by the sensor unit 103 of the charge / discharge device 100 and transmitted to the charge / discharge device terminal 200. Furthermore, each time the evaluation charge / discharge process is performed, at least one explanatory variable acquisition storage battery is disassembled and analyzed to acquire material analysis data and degradation information (S104). The material analysis data includes, for example, X-ray diffraction (XRD) spectra, nuclear magnetic resonance (NMR) spectra, electron spin resonance (ESR) spectra, scanning electron microscope (SEM) images, and ion chromatography (IC) spectra of the materials constituting the storage battery. The degradation information of the storage battery includes the degree of degradation of the storage battery's materials and components at the time of material analysis (after the evaluation charge / discharge protocol is performed). The degree of material deterioration is expressed as a percentage of the remaining functioning health of each material at the time of material analysis, with a new battery material (at the start of the life evaluation test) being 100%. The degree of component deterioration indicates the health of battery components, such as the positive electrode, negative electrode, separator, and electrolyte, at the end of their life. Examples of component deterioration include the degree of separator pore blockage (the rate of reduction in the number of pores) expressed as a percentage with a new battery material being 0%, separator breakage expressed as a binary evaluation of whether or not there is breakage, and the rate of reduction in separator strength expressed as a percentage with a new battery material being 100%. A method for calculating the degree of material deterioration includes comparing material analysis data of a storage battery in a new state with that at the time of material analysis, and calculating the percentage (mass %) of how well each material remains in a functioning state. Another method for calculating the degree of deterioration of components includes setting evaluation items and standards related to the health of each component in advance using thresholds or qualitatively, and calculating or determining the degree of deterioration of each component of the storage battery at the time of material analysis.Step S104 of acquiring charge / discharge data, material analysis data, and deterioration information is repeated until charge / discharge data, material analysis data, and deterioration information at the end of the battery's life (for example, when the capacity maintenance rate of the storage battery reaches 80% or when the temperature of the storage battery reaches its upper limit) are acquired.

[0098] In FIG. 6 , next, a plurality of explanatory variables are calculated (extracted) from the test condition data, charge / discharge data, and material analysis data (S105). The calculation of the explanatory variables is performed by the explanatory variable extraction unit 302 of the business operator server 300. The explanatory variables include test condition parameters selected from the test condition data, electrochemical parameters obtained by electrochemically analyzing the charge / discharge data, mathematical parameters obtained by mathematically processing the shape of the discharge curve in the charge / discharge process, and material analysis parameters obtained by processing the material analysis data. For details on obtaining the test condition parameters, electrochemical parameters, and mathematical parameters, see the explanation for FIG. 5 above. The obtained test condition parameters, electrochemical parameters, and mathematical parameters each become candidate explanatory variables when determining the "selected explanatory variables."

[0099] The material analysis parameters can be obtained by mathematically processing material analysis data of the storage battery, such as an X-ray diffraction (XRD) spectrum, a nuclear magnetic resonance (NMR) spectrum, an electron spin resonance (ESR) spectrum, a scanning electron microscope (SEM) image, or an ion chromatography (IC) spectrum of the material that constitutes the storage battery. Methods for obtaining the material analysis parameters from the material analysis data include the following processing steps (1) to (3).

[0100] Processing process (1) is a method for acquiring material analysis parameters from spectra such as XRD, NMR, ESR, and IC. First, for D samples, a matrix Y is obtained with D rows and E columns (D is the number of samples, D≧1; E is the number of elements, E≧2) in which the differences between the spectral data in a new state and the spectral data at the time of analysis are arranged as elements in the column direction. The number of battery samples D corresponds to the number of batteries used to acquire explanatory variables analyzed in step S104, and can be 1 or more, 2 or more, 5 or more, 10 or more, 20 or more, 50 or more, or 100 or more. Generally, the greater the number of samples D, the better the prediction accuracy of the trained model. The number E, which corresponds to the number of divisions of the spectral data along the horizontal axis, is approximately 10 to 1,000. Next, after applying mathematical constraints, matrix Y is decomposed into Y≈V×I (where matrix V is a coefficient matrix with D rows and F columns, and matrix I is a basis matrix with F rows and E columns). The mathematical constraints are set so that matrices V and I are uniquely determined. F is, for example, equal to or less than the number E of elements. Methods for matrix decomposition include principal component analysis and nonnegative matrix factorization. Then, F coefficients in matrix V (referred to as the "first principal component" to the "Cth principal component," starting from the first column) can be obtained as material analysis parameters. The obtained material analysis parameters become candidate explanatory variables when determining the "selected explanatory variables."

[0101] The details of the processing steps (2) and (3) are as explained in the "Explanatory Variable Extraction Unit" section of the "Business Operator Server" above. The acquired material analysis parameters become candidates for explanatory variables when determining the "selected explanatory variables."

[0102] Note that the order of conducting a predetermined life evaluation test and obtaining the evaluation results and deterioration information (S101 and S102) and the order of conducting a life evaluation test to obtain explanatory variables and calculating the explanatory variables (S103 to S105) may be reversed (S103 to S105 first, followed by S101 and S102). Furthermore, the predetermined life evaluation test and the life evaluation test to obtain explanatory variables may be performed in parallel (S101 and S102, and S103 to S105 in parallel).

[0103] 6, a database including explanatory variables and evaluation results is then created (S106). The database includes a plurality of explanatory variables calculated from the test condition data, charge / discharge data, and material analysis data described above, as well as evaluation results and degradation information from predetermined life evaluation tests linked to these explanatory variables, and is stored in the storage unit 303 of the business operator's server 300.

[0104] <Generation of Trained Model for Life Assessment Data> In FIG. 6 , finally, a trained model is generated and stored using the explanatory variables and the assessment results as training data (S107), thus completing the series of processes. Details of step S107 are similar to step S107 in FIG. 5 above, except that the explanatory variables include one or more material analysis parameters. The input variables are preferably a combination of explanatory variables selected from the plurality of explanatory variables to improve the prediction accuracy of the trained model and / or the interpretability of the predicted assessment results. The selection of explanatory variables is also similar to step S107 in FIG. 5 above, except that the explanatory variables include one or more material analysis parameters. Combinations of explanatory variables input to the trained model in FIG. 6 include, for example, one or more test condition parameters, one or more electrochemical parameters, and one or more material analysis parameters; and one or more test condition parameters, one or more electrochemical parameters, one or more mathematical parameters, and one or more material analysis parameters.

[0105] In this manner, a trained model can be generated that is trained to output a predicted evaluation result measured in a predetermined life evaluation test when multiple explanatory variables extracted from the charge / discharge data of the storage battery are input. The output of the generated trained model in FIG. 6 includes, as a predicted evaluation result, at least one selected from the post-test voltage retention rate of the storage battery, the post-test capacity retention rate, the life time to life, or the number of cycles to life, and further includes predicted deterioration information. The generated trained model is stored in the memory unit 303 of the business operator server 300.

[0106] 7 and 8 are flowcharts showing an example of the operation of the evaluation prediction process (evaluation result prediction method). Note that the flow of the operation described below is executed based on the evaluation result prediction program of the present disclosure stored in the storage unit 303 of the business operator's server 300, mainly by the explanatory variable extraction unit 302, the evaluation prediction unit 304, and the control unit 306 included in the processing unit of the business operator's server 300, in cooperation with each element of the evaluation result prediction device and system.

[0107] <Evaluation Prediction of Life Evaluation Test> FIG. 7 is a flowchart showing an example of the operation of the evaluation prediction process (evaluation result prediction method) using the trained model generated in FIG. 5 . First, one or more storage batteries 101 to be evaluated are connected in series or parallel to the power supply unit 102 of the charging / discharging device 100 to start a life evaluation test, and acquisition of test condition data begins (S201). A series of evaluation prediction processes is performed for each storage battery (for each storage battery). However, multiple storage batteries to be evaluated may be connected to one charging / discharging device 100, and the series of evaluation prediction processes may be performed in parallel for the multiple storage batteries. The test condition data is acquired (received) by the acquisition unit (receiving unit) 301 of the business operator server 300 from the charging / discharging device terminal 200 via the network N1. The conditions of the life evaluation test are the same as those of the specific life evaluation test to be predicted, except that an evaluation charge / discharge process is performed at the start of the life evaluation test and at regular intervals or every certain number of cycles.

[0108] 7 , a single charge / discharge process of the evaluation charge / discharge protocol is then performed on the storage battery at the start of the life evaluation test (in a brand new state), and charge / discharge data is acquired (S202). Next, after a certain period of time has passed or after a certain number of charge / discharge cycles have been completed, the evaluation charge / discharge process is performed again, and charge / discharge data is acquired (S203). The charge / discharge data is measured by the sensor unit 103 of the charge / discharge device 100 and transmitted to the acquisition unit (receiving unit) 301 of the business operator server 300 via the charge / discharge device terminal 200. Then, multiple explanatory variables are extracted from the test condition data and the charge / discharge data (S204). The explanatory variables are extracted by the explanatory variable extraction unit 302 of the business operator server 300. The multiple explanatory variables include test condition parameters selected from the test condition data and electrochemical parameters obtained by electrochemically analyzing the charge / discharge data, and optionally, one or more mathematical parameters obtained by mathematically processing the shape of the discharge curve in the charge / discharge process. Details of the test condition parameters, electrochemical parameters, and mathematical parameters are as described above in the "Explanatory Variable Extraction Unit" and in the "Creation of Database" section of FIG. 5. The explanatory variables may be pre-selected explanatory variables to improve the prediction accuracy of the trained model and / or improve the interpretability of the prediction evaluation results. Details of the pre-selected explanatory variables are as described above in the "Explanatory Variable Extraction Unit" and in the "Generation of Trained Model" section of FIG. 5.

[0109] In FIG. 7 , the extracted explanatory variables are then input into the trained model (S205), and the predicted evaluation results are obtained and the prediction accuracy is calculated (S206). The input of explanatory variables into the trained model, the acquisition of the predicted evaluation results, and the calculation of the prediction accuracy are performed by the evaluation prediction unit 304 of the business operator server 300. Details of the trained model and the predicted evaluation results are as described above in the sections "Memory Unit," "Learning Processing," and "Evaluation Prediction Unit." The longer the life evaluation test is continued, the more charge / discharge processes are completed and the more charge / discharge data is accumulated, resulting in higher prediction accuracy. Furthermore, the prediction accuracy of the input explanatory variables is improved by eliminating one or more explanatory variables that are multicollinear with other explanatory variables. Details of prediction accuracy are as described above in the section "Evaluation Prediction Unit."

[0110] <Charge / Discharge Control of Life Evaluation Test> In FIG. 7 , next, it is determined whether the prediction accuracy is equal to or greater than a threshold (S207). If the answer is No, i.e., if the prediction accuracy is less than the threshold, the life evaluation test continues until the next process of the evaluation charge / discharge protocol is performed (S208). The above steps (S203 to S208) are repeated until the answer to S207 becomes Yes. If the answer to S207 is Yes, the life evaluation test (evaluation charge / discharge protocol) is terminated (S209). The control of the charge / discharge device 100 based on the prediction accuracy is performed by the control unit 306 of the business operator server 300. Details of the control are as described above in the "Control Unit" section.

[0111] 7, finally, information (prediction information) including the prediction evaluation result and prediction accuracy is output (S210), and the series of processes is terminated. The prediction information is output by the output unit (transmission unit) 305 of the business operator server 300. Details of the output format are as explained above in the sections "Output Unit" and "Virtual Environment."

[0112] <Evaluation Prediction of Life Evaluation Test> Figure 8 is a flowchart showing an example of the operation of the evaluation prediction process (evaluation result prediction method) using the trained model generated in Figure 6. Figure 8 differs from the evaluation prediction process (evaluation result prediction method) of Figure 7 in that it involves the collection of material analysis data. First, a life evaluation test is started by connecting multiple storage batteries 101 to be evaluated in series or parallel to the power supply unit 102 of the charging / discharging device 100, and acquisition of test condition data is also started (S201). For details of step S201, please refer to the explanation of Figure 7.

[0113] In FIG. 8 , a single charge / discharge process of the evaluation charge / discharge protocol is then performed on the storage batteries (in a brand new state) at the start of the life evaluation test, and charge / discharge data is acquired. At least one storage battery is disassembled and material analysis data is acquired (S202). Next, after a certain period of time has passed or after a certain number of charge / discharge cycles have been completed, the evaluation charge / discharge process is performed again to acquire charge / discharge data. At least one storage battery is disassembled and material analysis data is acquired (S203). The charge / discharge data is measured by the sensor unit 103 of the charge / discharge device 100 and transmitted to the acquisition unit (receiving unit) 301 of the business operator server 300 via the charge / discharge device terminal 200. The material analysis data is transmitted from the business operator terminal 500 or the like to the acquisition unit (receiving unit) 301 of the business operator server 300 via the network N1. Then, multiple explanatory variables are extracted from the test condition data, charge / discharge data, and material analysis data (S204). The explanatory variables are extracted by the explanatory variable extraction unit 302 of the business operator server 300. The multiple explanatory variables include test condition parameters selected from the test condition data, electrochemical parameters obtained by electrochemically analyzing the charge / discharge data, and optionally, one or more mathematical parameters obtained by mathematically processing the shape of the discharge curve in the charge / discharge process. The multiple explanatory variables further include one or more material analysis parameters extracted from the material analysis data. Details of the test condition parameters, electrochemical parameters, mathematical parameters, and material analysis parameters are as described above in the "Explanatory Variable Extraction Unit" and the "Creation of Database" sections regarding FIG. 6. The explanatory variables may be pre-selected explanatory variables to improve the prediction accuracy of the trained model and / or improve the interpretability of the predicted evaluation results. Details of the pre-selected explanatory variables are as described above in the "Explanatory Variable Extraction Unit" and the "Generation of Trained Model" sections regarding FIG. 6.

[0114] In FIG. 8 , the extracted explanatory variables are then input into the trained model (S205), and the predicted evaluation results are obtained and the prediction accuracy is calculated (S206). The input of explanatory variables into the trained model, the acquisition of the predicted evaluation results, and the calculation of the prediction accuracy are performed by the evaluation prediction unit 304 of the business operator server 300. Details of the trained model and the predicted evaluation results are as described above in the sections "Memory Unit," "Learning Processing," and "Evaluation Prediction Unit." The longer the life evaluation test is continued, the more charge / discharge processes are completed and the more charge / discharge data is accumulated, resulting in higher prediction accuracy. Furthermore, the prediction accuracy of the input explanatory variables is improved by eliminating one or more explanatory variables that are multicollinear with other explanatory variables. Details of the prediction accuracy are as described above in the section "Evaluation Prediction Unit."

[0115] <Charge / Discharge Control of Life Evaluation Test> In FIG. 8 , next, it is determined whether the prediction accuracy is equal to or greater than a threshold (S207). If the answer is No, i.e., if the prediction accuracy is less than the threshold, the life evaluation test continues until the next process of the evaluation charge / discharge protocol is performed (S208). The above steps (S203 to S208) are repeated until the answer to S207 becomes Yes. If the answer to S207 is Yes, the life evaluation test (evaluation charge / discharge protocol) is terminated (S209). The control of the charge / discharge device 100 based on the prediction accuracy is performed by the control unit 306 of the business operator server 300. Details of the control are as described above in the "Control Unit" section.

[0116] In Fig. 8 , finally, information (prediction information) including the predicted evaluation result and prediction accuracy is output (S210), and the series of processes ends. The predicted information is output by the output unit (transmission unit) 305 of the business operator server 300. The predicted evaluation result as the output of the evaluation prediction process in Fig. 8 further includes predicted deterioration information at the end of the life. Details of the output format are as explained above in the sections "Output Unit" and "Virtual Environment".

[0117] 12 is a flowchart showing an example of the operation of the learning process of the trained model that outputs a predicted evaluation result including the energy density and power density measured in a predetermined charge / discharge protocol. Note that the flow of the operation described below is executed based on a program stored in the storage unit 303 of the business operator server 300, mainly by the explanatory variable extraction unit 302 and the learning processing unit 307 included in the processing unit of the business operator server 300, in cooperation with each element.

[0118] <Creation of Database of Energy Density and Power Density Data> First, one or more storage batteries 101 for collecting teacher data are connected in series or parallel to the power supply unit 102 of the charging / discharging device 100, and a predetermined charging / discharging protocol is executed (S301). The storage batteries for collecting teacher data are preferably of the same category as the storage battery to be evaluated (e.g., LiB, lead-acid battery, or nickel-metal hydride secondary battery), and more preferably, are storage batteries that share at least one of the materials (e.g., positive electrode active material, negative electrode active material, separator material, and electrolyte material) used in the components (e.g., positive electrode, negative electrode, separator, and electrolyte).

[0119] Next, evaluation results based on a predetermined charge / discharge protocol are acquired (S302). The evaluation results include energy density and power density obtained by processing charge / discharge data measured by the sensor unit 103 of the charge / discharge device 100. The acquired evaluation results are used as training data. Therefore, if a predetermined charge / discharge protocol has multiple measurement conditions for energy density and power density, all of them are basically acquired. For example, if the predetermined charge / discharge protocol is IEC 62660-1, the acquired evaluation results include (1) energy density at 25°C, (2) power density at 25°C and 20% SOC, (3) power density at 25°C and 50% SOC, (4) power density at 25°C and 80% SOC, (5) power density at -20°C and 50% SOC, (6) power density at 0°C and 50% SOC, and (7) power density at 45°C and 50% SOC. However, for some storage batteries, only some of these evaluation results may be acquired as training data.

[0120] Next, the evaluation charge / discharge protocol is implemented on one or more storage batteries connected in series or parallel (S303). The storage batteries on which the evaluation charge / discharge protocol is implemented are essentially the same individual storage batteries as the storage batteries on which the predetermined charge / discharge protocol was implemented. However, for purposes such as shortening the time required to create a database, the evaluation charge / discharge protocol may be implemented on other individual storage batteries from the same lot (i.e., multiple storage batteries manufactured at the same time using the same materials and under the same conditions) as the storage batteries on which the predetermined charge / discharge protocol was implemented. The charge / discharge data obtained by the evaluation charge / discharge protocol is later used to calculate explanatory variables that serve as input for the learning process. Therefore, if the evaluation charge / discharge protocol includes multiple single charge / discharge processes, essentially all of these are implemented. For example, when the evaluation charge / discharge protocol shown in Table 2 above is used, all of the first to fourth single charge / discharge processes are implemented. However, only some of these charge / discharge processes may be implemented for some storage batteries.

[0121] Next, charge / discharge data is acquired according to the evaluation charge / discharge protocol (S304). The charge / discharge data is measured by the sensor unit 103 of the charge / discharge device 100 and transmitted to the charge / discharge device terminal 200. Then, a plurality of explanatory variables are calculated (extracted) from the charge / discharge data (S305). The explanatory variables are calculated by the explanatory variable extraction unit 302 of the business operator server 300. The explanatory variables include both electrochemical parameters obtained by electrochemically analyzing the charge / discharge data and mathematical parameters obtained by mathematically processing the shape of the discharge curve in the charge / discharge process.

[0122] The electrochemical parameters include energy density and power density, and optionally, data on voltage relaxation behavior. The electrochemical parameters can be calculated (extracted) by electrochemically analyzing a single charge / discharge process of the evaluation charge / discharge protocol, and include energy density, power density, and voltage relaxation behavior. Details on the calculation of the electrochemical parameters are as described above in the "Explanatory Variable Extraction Section" section. For example, when the evaluation charge / discharge protocols shown in Table 2 above are used, the energy density, power density at SOC 80%, SOC 50%, and SOC 20%, and voltage relaxation behavior at each temperature are obtained from the first to fourth single charge / discharge processes. Each of the obtained electrochemical parameters becomes a candidate explanatory variable when determining the "selected explanatory variable."

[0123] Mathematical parameters can be extracted by mathematically processing the shape of a discharge curve in a charge / discharge process. The mathematical parameters include, for example, data obtained by mathematically processing a discharge curve in which the vertical axis represents the voltage of a storage battery and the horizontal axis represents any combination of time, current capacity, SOC, etc. FIG. 4 is a diagram illustrating an example of calculating mathematical parameters as explanatory variables. As illustrated in FIG. 4 , the mathematical parameters are obtained by matrix decomposition of a matrix representing the general shape of a discharge curve. First, for discharge curve data in which the horizontal axis represents the SOC value of a storage battery and the vertical axis represents the corresponding voltage (V) value, a matrix X with A rows and B columns is obtained in which B (B≧2) different SOC values ​​are arranged as columns and the corresponding voltage (V) values ​​of multiple storage batteries (the number of samples is A (A≧2)) are arranged as elements. The number B of SOC values ​​used may be approximately 10 to 1000. For example, as illustrated in FIG. 4 , when the SOC value is divided from 100% to 0% in 1% increments, B is 100%, 99%, 98%, ..., 2%, 1%, and 0%, for a total of 101 values. The number of battery samples A is, for example, 10 or more, 50 or more, 100 or more, 500 or more, or 1000 or more. The larger the number of samples A, the more the prediction accuracy of the trained model improves. Next, after applying mathematical constraints, matrix X is decomposed into X≈W×H (where matrix W is a coefficient matrix with A rows and C columns, and matrix H is a basis matrix with C rows and B columns). The mathematical constraints are set so that matrices W and H are uniquely determined. C is, for example, equal to or less than the number B of SOC values. Methods of matrix decomposition that can be used include principal component analysis and nonnegative matrix factorization. The C coefficient matrices in the matrix W (referred to as the "first principal component" to the "Cth principal component" in order from the first column) are each candidates for explanatory variables when determining the "selected explanatory variables."

[0124] Note that the execution of the predetermined charge / discharge protocol and acquisition of the evaluation results (S301 and S302) and the execution of the evaluation charge / discharge protocol and calculation of the explanatory variables (S303 to S305) may be performed in the reverse order (S303 to S305 first, and then S301 and S302). Furthermore, when multiple storage batteries from the same lot are used as described above, the predetermined charge / discharge protocol and the evaluation charge / discharge protocol may be performed in parallel (S301 and S302, and S303 to S305 in parallel).

[0125] Next, a database including explanatory variables and evaluation results is created (S306). The database includes multiple explanatory variables calculated from the charge / discharge data according to the above-described evaluation charge / discharge protocol and the associated evaluation results according to a predetermined charge / discharge protocol. The database is stored in the storage unit 303 of the business operator server 300. For example, when the evaluation charge / discharge protocol shown in Table 2 above is used and B = C = 101 as illustrated in FIG. 4, a single charge / discharge process generates five electrochemical parameters (energy density, power density at 80% SOC, power density at 50% SOC, power density at 20% SOC, and voltage relaxation behavior) and 101 mathematical parameters (first principal component to 101st principal component) for each storage battery. In other words, when all of the first to fourth single charge / discharge processes are performed, a total of 424 parameters are generated (4 temperatures x 106). These parameters are stored in a database as explanatory variables (input variables) that provide evaluation results, which are response variables (output variables).

[0126] <Generation of Trained Model for Energy Density and Power Density Data> Finally, a trained model is generated and stored using the explanatory variables and the evaluation results as training data (S307), and the series of processes is terminated. The trained model can be generated by pre-training a machine learning model using supervised learning such as ridge regression, support vector regression, or neural network. As input variables, data obtained by standardizing the above-mentioned explanatory variables (processing to set the average of the data to 0 and the standard deviation to 1) is used. Note that the evaluation results as output variables may also be standardized. Next, the machine learning model is trained on the input variables and the corresponding output variables (evaluation results), and the prediction accuracy of the model is calculated. As an index of prediction accuracy, the coefficient of determination (R ２ The following can be used: the root mean square error (RMSE), the mean absolute error (MAE), the mean absolute percentage error (MAPE), etc. When verifying the prediction accuracy, it is preferable to perform cross-validation. Then, the hyperparameters are adjusted so that the prediction accuracy of the machine learning model is maximized.

[0127] The input variables are preferably a combination of explanatory variables selected from the plurality of explanatory variables so as to improve the prediction accuracy of the trained model and / or the interpretability of the prediction evaluation results. More specifically, the input variables may be explanatory variables selected by eliminating one or more parameters that are multicollinear with other parameters, rather than using all of the parameters obtained in the calculation of the explanatory variables (S305). Methods for selecting explanatory variables include, for example, methods using Boruta, recursive feature reduction, genetic algorithms (GA), etc. Boruta is a variable selection method based on the variable importance of random forests (RF). The method selects explanatory variables by adding dummy variables to the explanatory variables, constructing a random forest model, calculating the importance of each variable, and removing variables less important than the dummy variables. Recursive feature reduction is a method in which a predictive model is created using all explanatory variables, the least important explanatory variable is removed, a new predictive model is created using the remaining explanatory variables, and the least important explanatory variable is removed again, until a pre-specified number of explanatory variables is reached. Genetic algorithms (GAs) are a method for solving optimization problems, considering whether each explanatory variable should be used or not (1 or 0), and determining which combination of explanatory variables maximizes accuracy. From the perspective of prediction accuracy and interpretability, it is preferable to try all of these methods and select the one that provides the best prediction accuracy and interpretability. For example, Table 4 below shows examples of explanatory variables selected using a genetic algorithm (GA) for a total of 424 parameters obtained from the evaluation charge / discharge protocol shown in Table 2 above. These selected explanatory variables correspond to the "preselected explanatory variables" (Table 3 above) in the operation of the evaluation result prediction device and system, and evaluation prediction method and program of the present disclosure. The combinations of explanatory variables and prediction accuracy shown in Table 4 below are merely examples.

[0128]

[0129] In this way, a trained model can be generated that is trained to output a predicted evaluation result including the energy density and power density measured using a predetermined charge / discharge protocol when multiple explanatory variables extracted from the charge / discharge data of the storage battery are input. The generated trained model is stored in the storage unit 303 of the business operator server 300.

[0130] 13 is a flowchart showing an example of the operation of the evaluation prediction process (evaluation result prediction method). Note that the flow of the operation described below is executed based on the evaluation result prediction program of the present disclosure stored in the storage unit 303 of the business operator server 300, mainly by the explanatory variable extraction unit 302, the evaluation prediction unit 304, and the control unit 306 included in the processing unit of the business operator server 300, in cooperation with each element of the evaluation result prediction device and system.

[0131] <Evaluation and Prediction of Energy Density and Power Density> First, one or more storage batteries 101 to be evaluated are connected in series or parallel to the power supply unit 102 of the charging / discharging device 100, and a first single charging / discharging process of the evaluation charging / discharging protocol is performed (S401). A series of evaluation and prediction processes is performed for one storage battery (for each storage battery). However, a plurality of storage batteries to be evaluated may be connected to a single charging / discharging device 100, and a series of evaluation and prediction processes may be performed in parallel for the plurality of storage batteries.

[0132] Next, charge / discharge data is acquired (S402). The charge / discharge data is measured by the sensor unit 103 of the charge / discharge device 100 and transmitted to the charge / discharge device terminal 200. Then, multiple explanatory variables are extracted from the charge / discharge data (S403). The explanatory variables are extracted by the explanatory variable extraction unit 302 of the business operator server 300. The multiple explanatory variables include electrochemical parameters obtained by electrochemically analyzing the charge / discharge data, and optionally further include one or more mathematical parameters obtained by mathematically processing the shape of the discharge curve in the charge / discharge process. Details of the electrochemical parameters and mathematical parameters are as described above in the "Explanatory Variable Extraction Unit" and "Creation of Database" sections. The explanatory variables may be pre-selected explanatory variables to improve the prediction accuracy of the trained model and / or improve the interpretability of the prediction evaluation results. Details of the pre-selected explanatory variables are as described above in the "Explanatory Variable Extraction Unit" and "Generation of Trained Model" sections.

[0133] Next, the extracted explanatory variables are input into the trained model (S404), and the predicted evaluation results are obtained (S405). The input of explanatory variables into the trained model and the acquisition of the predicted evaluation results are performed by the evaluation prediction unit 304 of the business operator server 300. Details of the trained model and the predicted evaluation results are as explained above in the sections "Memory unit," "Learning process," and "Evaluation prediction unit."

[0134] Next, the prediction accuracy of the predicted evaluation result is calculated (S406). The calculation of the prediction accuracy is also performed by the evaluation prediction unit 304 of the business operator server 300. The more charging / discharging processes are completed and the more charging / discharging data is accumulated, the higher the prediction accuracy becomes. Furthermore, the prediction accuracy of the input explanatory variables can be improved by excluding one or more explanatory variables that are multicollinear with other explanatory variables (Tables 2 and 3). Details of the prediction accuracy are as explained above in the "Evaluation Prediction Unit" section.

[0135] <Control of Charging and Discharging Energy Density and Power Density> Next, it is determined whether the prediction accuracy is equal to or greater than a threshold value or whether all processes have been completed (S407). If the answer is No, i.e., the prediction accuracy is less than the threshold value and all processes have not been completed, the next process in the evaluation charge / discharge protocol is executed (S408). The above steps (S402 to S408) are repeated until the answer in S407 is Yes. If the answer in S407 is Yes, the evaluation charge / discharge protocol is terminated (S409). The control of the charge / discharge device 100 based on the prediction accuracy is performed by the control unit 306 of the business operator server 300. Details of the control are as described above in the "Control Unit" section.

[0136] Finally, information (prediction information) including the prediction evaluation result and prediction accuracy is output (S410), and the series of processes is terminated. The prediction information is output by the output unit (transmission unit) 305 of the business operator server 300. Details of the output format are as explained above in the sections "Output Unit" and "Virtual Environment".

[0137] <<Modifications>> The above describes, as an example of an embodiment, a case in which a business operator server in an evaluation result prediction system functions as an “evaluation result prediction device” that predicts evaluation results measured in a predetermined life evaluation test and / or predicts the energy density and power density of a storage battery sequentially measured using a predetermined charge / discharge protocol. However, in the evaluation result prediction system of the present disclosure, any device constituting the system may function as an “evaluation result prediction device” as long as the evaluation prediction process of the present disclosure can be performed as an entire system interconnected via a network. Each device (functional unit) constituting the evaluation result prediction system, i.e., an acquisition device (acquisition unit), explanatory variable extraction device (explanatory variable extraction unit), storage device (storage unit), evaluation prediction device (evaluation prediction unit), output device (output unit), control device (control unit), and learning processing device (learning processing unit), may be configured as a single device (a single device may have all the functional units) or may be configured as multiple devices interconnected via a network. When the evaluation result prediction system is configured as multiple devices, each device may have one or more functional units, and the same functional unit may be implemented in multiple devices.

[0138] <Variation 1> Fig. 9 is a diagram illustrating an example of the evaluation result prediction system 1 according to the embodiment of the present disclosure of Fig. 1. The evaluation result prediction system 1 includes a charging / discharging device 100, a charging / discharging device terminal 200 as an "evaluation result prediction device" communicatively connected to the charging / discharging device 100, a business operator server 300 communicatively connected to the charging / discharging device terminal 200 via a business operator's private network N1, and a virtual environment 400 constructed on a network N2 and communicatively connected to the business operator server 300. The evaluation result prediction system 1 is accessible from terminal devices such as a business operator terminal 500 and a customer terminal 600, which are indicated by dashed lines. The business operator terminal 500 is communicatively connected to the charging / discharging device terminal 200 and the business operator server 300 via the business operator's private network N1, and is communicatively connected to the virtual environment 400 and the customer terminal 600 via the network N2. The customer terminal 600 is communicatively connected to the virtual environment 400 and the business operator terminal 500 via the network N2. The evaluation result prediction system 1, mainly through the charging / discharging device terminal 200, can work in cooperation with each element to predict the evaluation results that would be obtained if a specified life evaluation test and / or a specified charge / discharge protocol were performed on the storage battery to be evaluated, based on the test condition data for the life evaluation test and the charge / discharge data obtained by performing charging / discharging based on the evaluation charge / discharge protocol on the storage battery to be evaluated.

[0139] The charging / discharging device terminal 200 in FIG. 9 functions as an "evaluation result prediction device" by including an acquisition unit 201, an explanatory variable extraction unit 202, a memory unit 203 that stores the reflected trained model, etc., an evaluation prediction unit 204, and an output unit (transmission unit of the transmission / reception unit) 205. The acquisition unit 201 acquires test condition data from the charging / discharging device 100. The acquisition unit 201 also acquires charge / discharge data of the storage battery based on the evaluation charge / discharge protocol directly from the sensor unit 103 of the charging / discharging device 100. The acquisition unit 201 also acquires (receives) material analysis data of the storage battery from the business operator terminal 500 or the like via the network N1. The explanatory variable extraction unit 202 extracts multiple explanatory variables from the test condition data, the charge / discharge data, and the material analysis data. And / or the explanatory variable extraction unit 202 extracts multiple explanatory variables including electrochemical parameters, including energy density and power density, from the charge / discharge data. The memory unit 203 stores a reflected trained model that has been trained to output a predicted evaluation result measured in a predetermined life evaluation test when multiple explanatory variables are input, and / or that has been trained to output a predicted evaluation result including energy density and power density measured in a predetermined charge / discharge protocol when multiple explanatory variables are input. The "reflected" trained model is a trained model stored in the memory unit 203 of the charging / discharging device terminal 200 by receiving a trained model generated by the business operator server 300 via the network N1. If an update to the trained model is made on the business operator server 300 side, the update is reflected as needed. The evaluation / prediction unit 204 inputs multiple explanatory variables into the trained model to obtain a predicted evaluation result and calculate the prediction accuracy of the predicted evaluation result. The output unit (transmitting unit of the transceiver unit) 205 outputs (transmits) prediction information including the predicted evaluation result and prediction accuracy to the acquisition unit (receiving unit) 301 of the business operator server 300 via the network N1. In addition, the output unit (transmitting unit of the transceiver unit) 205 outputs (transmits) data necessary for constructing a database, such as the evaluation results of a specified life evaluation test and / or a specified charge / discharge protocol, as well as the charge / discharge data of the evaluation charge / discharge protocol and calculated explanatory variables, to the acquisition unit (receiving unit) 301 of the operator server 300 via the network N1.The charging / discharging device terminal 200 further includes a control unit 206, and can control the charging / discharging device 100 based on the prediction accuracy.

[0140] The business operator server 300 in FIG. 9 includes an acquisition unit (receiving unit) 301, a storage unit 303 that stores a trained model, a database, and the like, an output unit (transmitting unit) 305, and a learning processing unit 307. The acquisition unit (receiving unit) 301 acquires (receives) prediction information and data necessary for constructing a database from the output unit (transmitting unit of the transceiver unit) 205 of the charging / discharging device terminal 200 via the network N1, and stores the data in the storage unit 303. The output unit (transmitting unit) 305 outputs (transmits) the prediction information to the virtual environment 400 via the network N2 in real time. The learning processing unit 307 can generate and re-train a trained model based on the constructed database. The generated trained model is stored in the storage unit 303 and transmitted from the output unit (transmitting unit) 305 via the network N1 to the receiving unit (receiving unit of the transceiver unit) 205 of the charging / discharging device terminal 200, and is reflected in the storage unit 203 of the charging / discharging device terminal 200.

[0141] 9 , business operator terminal 500 is communicatively connected to charging / discharging device terminal 200 and business operator server 300 via a business operator's private network N1. Business operator terminal 500 is also communicatively connected to virtual environment 400 and customer terminal 600 via network N2. Customer terminal 600 is communicatively connected to virtual environment 400 via network N2 and can display the GUI of virtual environment 400.

[0142] For details of the functional units and processes in Figure 9, the descriptions of the functional units and processes explained with reference to Figures 1 to 8 are cited. In the embodiment of Figure 9, the charging / discharging device terminal 200 that performs the rating prediction process and the business operator server 300 that generates the trained model and stores the data are separated, which has the advantage of enabling the distribution of hardware resources for calculations, data storage, and the like.

[0143] <Modification 2> Fig. 10 is a diagram showing a schematic configuration of an evaluation result prediction system 1 according to an embodiment of the present disclosure. The evaluation result prediction system 1 includes a charging / discharging device 100, a charging / discharging device terminal 200 communicatively connected to the charging / discharging device 100, a business operator server 300 communicatively connected to the charging / discharging device terminal 200 and the shared server 700 via a business operator's private network N1, and a shared server 700 communicatively connected to the charging / discharging device terminal 200 and the business operator server 300 via the network N1. The evaluation result prediction system 1 is accessible from terminal devices such as a business operator terminal 500 and a customer terminal 600, which are indicated by dashed lines. The business operator terminal 500 is communicatively connected to the charging / discharging device terminal 200, the business operator server 300, and the shared server 700 via the business operator's private network N1, and is communicatively connected to the shared server 700 and the customer terminal 600 via the network N2. The customer terminal 600 is communicatively connected to the shared server 700 and the business operator terminal 500 via the network N2.

[0144] FIG. 11 is a diagram illustrating an example of the evaluation result prediction system 1 according to the embodiment of the present disclosure illustrated in FIG. 10 . The shared server 700 in FIG. 11 is an information processing terminal positioned above the charging / discharging terminal device 200 in terms of the control system, and more specifically, is, for example, a SCADA terminal. The shared server 700 functions as an "evaluation result prediction device" by including an acquisition unit (receiving unit) 701, an explanatory variable extraction unit 702, a memory unit 703 that stores the reflected trained model, etc., an evaluation prediction unit 704, and an output unit (GUI) 705. The acquisition unit (receiving unit) 701 acquires test condition data from the charging / discharging device terminal 200 via the network N1. The acquisition unit (receiving unit) 701 also acquires (receives) charging / discharging data of the storage battery based on the evaluation charging / discharging protocol from the transmitting / receiving unit (205) of the charging / discharging device terminal 200 via the network N1. Furthermore, the acquisition unit (receiving unit) 701 acquires (receives) material analysis data of the storage battery from the business operator terminal 500, etc., via the network N1. The explanatory variable extraction unit 702 extracts multiple explanatory variables from the test condition data, charge / discharge data, and material analysis data. And / or, the explanatory variable extraction unit 702 extracts multiple explanatory variables, including electrochemical parameters such as energy density and power density, from the charge / discharge data. The memory unit 703 stores a reflected trained model that has been trained to output predicted evaluation results measured in a predetermined life evaluation test and / or to output predicted evaluation results including energy density and power density measured in a predetermined charge / discharge protocol when multiple explanatory variables are input. The "reflected" trained model is a trained model generated by the business operator server 300 and stored in the memory unit 703 of the shared server 700 by receiving the trained model via the network N1. If the trained model is updated on the business operator server 300 side, the update is reflected as soon as it is received. The evaluation prediction unit 704 obtains a predicted evaluation result by inputting multiple explanatory variables into the trained model, and calculates the prediction accuracy of the predicted evaluation result. The output unit (GUI) 705 outputs (displays) prediction information including the predicted evaluation result and the prediction accuracy in the form of a GUI. The shared server 700 further includes a control unit 706, and can control the charging / discharging device 100 based on the prediction accuracy.

[0145] The business operator server 300 in FIG. 11 includes an acquisition unit (receiving unit) 301, a storage unit 303 that stores a trained model, a database, and the like, an output unit (transmission unit) 305, and a learning processing unit 307. The acquisition unit (receiving unit) 301 acquires (receives) data necessary for constructing a database, such as evaluation results of a predetermined life evaluation test and / or a predetermined charge / discharge protocol, charge / discharge data of an evaluation charge / discharge protocol, and calculated explanatory variables, from the transmission / reception unit 205 of the charging / discharging device terminal 200 via the network N1, and stores the data in the storage unit 303. The learning processing unit 307 can generate and retrain a trained model based on the constructed database. The generated trained model is stored in the storage unit 303 and is transmitted from the output unit (transmission unit) 305 to the acquisition unit (receiving unit) 701 of the shared server 700 via the network N1, where it is reflected in the storage unit 703 of the shared server 700.

[0146] 11 , business operator terminal 500 is communicatively connected to charging / discharging device terminal 200, business operator server 300, and shared server 700 via a business operator's private network N1. Business operator terminal 500 is also communicatively connected to shared server 700 and customer terminal 600 via network N2. Customer terminal 600 is communicatively connected to shared server 700 via network N2 and can display the GUI of shared server 700.

[0147] For details of the functional units and processes in Figure 11, the descriptions of the functional units and processes explained with reference to Figures 1 to 8 are cited. In the aspect of Figure 11, the shared server 700 that performs the rating prediction process and the business operator server 300 that generates the trained model and stores the data are separated, which has the advantage of enabling the distribution of hardware resources for calculations, data storage, and the like.

[0148] 14 shows an example of processing in the operable factor prediction method of the present disclosure. The operable factor prediction method of the present disclosure is a method for predicting operable factors, which are design factors of a storage battery that affect the lifespan of the storage battery. Note that the processing flow described below is based on an operable factor prediction program of the present disclosure stored in a memory unit or storage device in the operable factor prediction device or system, and can be executed mainly by the acquisition unit (acquisition device), deterioration level calculation unit (deterioration level calculation device), memory unit (storage device), simplified association model generation unit (simplified association model generation device), and output unit (output device) in the operable factor prediction device or system, in cooperation with each element of the operable factor prediction device or system.

[0149] In Figure 14, the method includes a step (S501) of acquiring predicted deterioration information after deterioration of the storage battery; a step (S502) of acquiring an association model in which causal relationships are previously associated between a life determining event that serves as a basis for determining the life of the storage battery, a plurality of deterioration events, and a plurality of candidate operable factors; a step (S503) of calculating the degree of deterioration for each of the life determining event and the plurality of deterioration events from the acquired predicted deterioration information; a step (S504) of simplifying the association model by, for example, excluding the deterioration events with a low degree of deterioration; and a step (S506) of outputting the candidate operable factors that have a causal relationship with the life determining event in the simplified association model as operable factors.

[0150] Generally, degradation events that occur with battery degradation include a wide variety of events, such as charge / discharge characteristic degradation events, such as an increase in resistance components in charge / discharge data; electrochemical analysis degradation events, such as an increase in ohmic resistance, an increase in reaction resistance, and an increase in diffusion resistance; and material analysis degradation events, such as SEI formation, an increase in current collection resistance, and separator clogging. Each degradation event has a complex causal relationship with the life-determining events that serve as the basis for determining the life of the battery. Furthermore, there are a huge number of operable factors, which are design factors for the battery that can improve these degradation events, and these factors have complex causal relationships with the life-determining events and multiple degradation events. Therefore, it is difficult to predict, from predicted degradation information after battery degradation, which specific operable factors should be manipulated to efficiently improve the life of the battery. In this regard, according to the method disclosed herein, an association model is prepared that pre-associates the causal relationships between a life-determining event, multiple degradation events, and multiple candidate operable factors. The association model is then simplified based on the degree of degradation, eliminating degradation events with a small degree of degradation, i.e., degradation events with a small range of possible improvements (hereinafter referred to as "improvement margin"), and degradation events that are considered to have a low or no causal relationship with the life-determining event. This allows for the identification of major degradation events that have a direct or indirect correlation with the life-determining event, making it easy to extract operable factors that affect the lifespan. Therefore, the method disclosed herein can contribute to the efficient improvement of the lifespan of storage batteries.

[0151] Fig. 15 shows a modified example of processing in the manipulable factor prediction method of the present disclosure. Note that the processing flow of Fig. 15 can also be executed based on the manipulable factor prediction program of the present disclosure stored in a memory unit or storage device in the manipulable factor prediction device or system, mainly by the acquisition unit (acquisition device), deterioration degree calculation unit (deterioration degree calculation device), memory unit (storage device), simplified association model generation unit (simplified association model generation device), influence degree acquisition unit (influence degree acquisition device), and output unit (output device) in the manipulable factor prediction device or system, in cooperation with the manipulable factor prediction device or system.

[0152] As shown in FIG. 15 , the operable factor prediction method may further include a step (S505) of acquiring the influence of a plurality of deterioration events on a life-determining event. In the influence acquisition step (S505), the deterioration level of the life-determining event and the deterioration levels of the plurality of deterioration events are input to a trained model that is trained to output an influence level representing the improvement effect on the deterioration level of the life-determining event when one or more of the plurality of deterioration events are improved. This allows the influence levels corresponding to one or more of the plurality of deterioration events to be acquired. In this aspect, in the output step (S506), one or more operable factor candidates that have a direct or indirect causal relationship with one or more deterioration events whose influence level is equal to or greater than a threshold value in the simplified association model can be output as operable factors affecting the life of the storage battery. Here, the “influence level” represents the degree of contribution to the life-improvement effect, i.e., the extent to which the deterioration level of the life-determining event can be improved when the deterioration level of the deterioration event is improved. 14, the embodiment of Fig. 15 uses a simplified correlation model based on the deterioration level, and therefore precludes operable factor candidates whose deterioration level is lower than a threshold and candidate operable factors that are considered to have no causal relationship with the lifespan determining events. By considering the influence level in addition to the deterioration level, it is possible to easily predict operable factors that have a large margin for improvement and have a high improvement effect on the deterioration level of the lifespan determining events, which can greatly contribute to the efficient improvement of the lifespan of the storage battery.

[0153] <Step of Acquiring Predicted Deterioration Information> In the step (S501) of acquiring predicted deterioration information, at least one piece of predicted deterioration information selected from the group consisting of predicted charge / discharge characteristics, predicted electrochemical analysis data, and predicted material analysis data after deterioration of the storage battery is acquired. Here, a "deteriorated" storage battery refers to a storage battery in an arbitrary state of deterioration when a new storage battery is repeatedly charged and discharged. For example, a deteriorating storage battery may be a storage battery at the end of its life when a predetermined life evaluation test is conducted on the storage battery. More specifically, examples of the predetermined life evaluation test include a storage battery storage test and a cycle test as specified in IEC 62660-1 (Performance Test) standardized by the International Electrotechnical Commission (IEC).

[0154] Examples of predicted charge / discharge characteristics include the energy density, power density, capacity, and charge / discharge efficiency of the storage battery at the end of its life. It is preferable that the predicted deterioration information include these charge / discharge characteristics in order to more accurately calculate the degree of deterioration of charge / discharge characteristic deterioration events. Examples of predicted electrochemical analysis data include charge curve analysis (CCA), discharge curve analysis, constant current intermittent titration (GITT), and electrochemical impedance (EIS) data of the storage battery at the end of its life. It is preferable that the predicted deterioration information include these electrochemical analysis data in order to more accurately calculate the degree of deterioration of electrochemical analysis deterioration events and material analysis deterioration events. Examples of predicted material analysis data include nuclear magnetic resonance (NMR) data, ion chromatography (IC) data, scanning electron microscope (SEM) observation data, X-ray diffraction (XRD) data, electron spin resonance (ESR) data, X-ray CT image data, and air permeability measurement data of the materials constituting the storage battery at the end of its life. It is preferable that the predicted deterioration information includes these material analysis data in order to more accurately calculate the deterioration levels of electrochemical analytical deterioration events and material analytical deterioration events.

[0155] One method for acquiring predicted deterioration information is to use a trained model (hereinafter referred to as "trained model I") that has been trained to output predicted evaluation results including predicted deterioration information that would be obtained if a life evaluation test were conducted when test condition data for the life evaluation test, as well as multiple explanatory variables extracted from charge / discharge data and material analysis data during the life evaluation test, are input. The multiple explanatory variables can be input into the trained model I to acquire predicted deterioration information. This allows the results of the life evaluation test of the storage battery to be predicted from data before the test is completed, thereby shortening the time required for the life evaluation test of the storage battery. As described below, the trained model I is stored in a memory unit in the operable factor prediction device or a storage device in the operable factor prediction system, and can be acquired from the memory unit or storage device.

[0156] Examples of the test condition data include test condition parameters such as the temperature, current value, voltage (e.g., upper limit voltage, lower limit voltage), and charge / discharge rest time of the storage battery. The test condition parameters can be used as they are as explanatory variables of the trained model I without any processing. However, the test condition data may be mathematically processed and used as the test condition parameters.

[0157] The charge / discharge data can be obtained, for example, by charging and discharging the storage battery based on a predetermined evaluation charge / discharge protocol during a life evaluation test. Then, electrochemical parameters and mathematical parameters calculated from the charge / discharge data can be used as explanatory variables of the trained model I. The evaluation charge / discharge protocol can be, for example, a charge / discharge protocol that includes measuring energy density and power density during a life evaluation test. The evaluation charge / discharge protocol preferably includes measuring the energy density and power density in a single charge / discharge process, and the single charge / discharge process includes discharging the storage battery at a predetermined temperature and a predetermined current value to measure the energy density, and temporarily increasing a maximum current value I when the charge rate reaches a target charge rate during the discharge process. ｍａｘand measuring the power density at the target charging rate, and then returning the current to the predetermined value again, each time one or more target charging rates are reached, thereby measuring the energy density and the power density in a single discharge process.

[0158] 2A is a diagram showing an example of a single discharge process in the evaluation charge / discharge protocol. The single discharge process in FIG. 2A includes measuring the energy density of a storage battery adjusted to 100% SOC by discharging the storage battery at a current value of 1 C until the SOC drops from 100% to 0% (Evaluation A). In addition, during the process of performing Evaluation A, when the SOC of the storage battery reaches 80%, a maximum current value I of 2 C or more is measured. ｍａｘ for 10 seconds to measure the power density at an SOC of 80%, and then suspend charging and discharging for 1 minute, after which discharging is continued again at a current value of 1 C (Evaluation B). Similar operations are performed when the SOC of the storage battery reaches 50% and when it reaches 20%, and the power densities at SOCs of 50% and 20% are measured (Evaluation C and Evaluation D).

[0159] The evaluation charge / discharge protocol may include first performing the single discharge process illustrated in FIG. 2(a) on a new storage battery at the start of a life evaluation test, and then performing the discharge process every fixed period (X') shorter than the storage period (X) of the storage battery in the specified life evaluation test, or every fixed number of charge / discharge cycles (Y') shorter than the number of charge / discharge cycles (Y) performed in the specified life evaluation test. By performing the single discharge process at the start of the life evaluation test, charge / discharge data of the new storage battery can be obtained, and electrochemical parameters in the new state, such as energy density, power density, and discharge capacity, can be calculated from the obtained charge / discharge data. For example, when the evaluation results of a storage battery storage test (X = 28 days or 42 days) according to IEC 62660-1 are to be predicted, the evaluation charge / discharge protocol may include performing the single discharge process illustrated in FIG. 2(a) at 45°C, for example, every 14 days, every 7 days, every 4 days, every 2 days, or every day. When the evaluation results of a cycle test of a storage battery according to IEC 62660-1 are to be predicted, the evaluation charge / discharge protocol may include performing a single discharge process, as illustrated in FIG. 2(a), at 45°C, for example, every 14 days, 7 days, 4 days, 2 days, or every day. Alternatively, the evaluation charge / discharge protocol may include performing the single discharge process at 45°C, for example, every 100 cycles, 50 cycles, 25 cycles, 20 cycles, or 10 cycles. This allows predicted evaluation results and prediction accuracy to be obtained for a certain period or every certain number of charge / discharge cycles. Once the user's desired prediction accuracy is achieved, the evaluation charge / discharge protocol can be terminated (i.e., the life evaluation test can be terminated), eliminating the need for subsequent life evaluation tests. This significantly reduces the time required for life evaluation testing. For example, FIG. 2(b) shows an example of an evaluation charge / discharge protocol that includes performing a single discharge process every certain number of charge / discharge cycles. FIG. 2(b) shows that the single discharge process of FIG. 2(a) described above is first performed on a new storage battery at the start of a life evaluation test, and then performed every time the number of charge / discharge cycles (e.g., Y′=100) is less than the number of charge / discharge cycles performed in the cycle test of a storage battery in IEC 62660-1.

[0160] The electrochemical parameters may include, for example, data on at least one selected from energy density, power density, voltage relaxation behavior, and capacity retention rate. The electrochemical parameters can be extracted by electrochemically analyzing the charge / discharge process of the evaluation charge / discharge protocol. For example, the energy density can be calculated from the cumulative current capacity (Wh) when the storage battery is discharged from 100% SOC to 0% and the battery weight (kg) or battery volume (L) using the following formula: gravimetric energy density (Wh / kg) = cumulative current capacity (Wh) / battery weight (kg), or the following formula: volumetric energy density (Wh / L) = cumulative current capacity (Wh) / battery volume (L). The power density can be calculated by temporarily calculating the maximum current value I ｍａｘ The voltage drop (V) after discharge when (A) is output, and the battery weight (kg) or battery volume (L) are used to calculate the weight power density (W / kg) = voltage drop (V) × current value I ｍａｘ (A) / battery weight (kg), or the following formula: volume power density (W / L) = voltage drop (V) × current value I ｍａｘ (A) / battery volume (L). The battery weight (kg) or battery volume (L) is obtained separately from the charge / discharge data. The voltage relaxation behavior is calculated based on the maximum current value I ｍａｘ The behavior of overvoltage mitigation when charging and discharging are temporarily stopped after the output is parameterized, and is expressed as, for example, the following formula: Voltage relaxation = Voltage after charging and discharging are stopped for 1 minute - Maximum current value I ｍａｘ The capacity retention rate can be calculated from the discharge capacity of the storage battery in a new state (at the start of the life evaluation test) under the conditions for implementing the life evaluation test (e.g., the integrated current capacity when discharging from SOC 100% to 0%) and the discharge capacity of the storage battery at a measurement point during the life evaluation test (e.g., the point at which a single charge / discharge process of the evaluation charge / discharge protocol is performed) using the following formula: capacity retention rate = (discharge capacity of the storage battery at the measurement point during the life evaluation test / discharge capacity of the storage battery at the start of the life evaluation test) × 100.

[0161] The mathematical parameters can be extracted by mathematically processing the shape of the discharge curve in the charge / discharge process. The mathematical parameters include, for example, data obtained by mathematically processing a discharge curve in which the vertical axis represents the voltage of the storage battery and the horizontal axis represents any combination of time, current capacity, SOC, etc. More specifically, as the mathematical parameters, a matrix X with 1 row and B columns is first obtained, in which voltage (V) values ​​corresponding to B (B≧2) different SOC values ​​are arranged as elements in the column direction for discharge curve data in which the SOC value of the storage battery to be evaluated is represented on the horizontal axis and the corresponding voltage (V) value is represented on the vertical axis. The number B of SOC values ​​used is approximately 10 to 1000, corresponding to the number B of SOC values ​​used in generating the trained model. For example, as illustrated in FIG. 4 , if the SOC value is divided in 1% increments from 100% to 0%, B is a total of 101, including 100%, 99%, 98%, ..., 2%, 1%, and 0%. The number of rows in matrix X is 1 because the rating prediction process is performed for each storage battery (for each storage battery). Next, after applying mathematical constraints, matrix X is decomposed into X≈W×H (where matrix W is a coefficient matrix with 1 row and C columns, and matrix H is a basis matrix with C rows and B columns). The mathematical constraints are set so that matrices W and H are uniquely determined. C corresponds to C used when generating the trained model and is, for example, equal to or less than the number B of SOC values. Methods for matrix decomposition include principal component analysis and nonnegative matrix factorization. Then, one or more explanatory variables can be selected from the C coefficients in matrix W (referred to as the "first principal component" to the "Cth principal component," starting from the first column).

[0162] The material analysis data can be obtained, for example, by measuring nuclear magnetic resonance (NMR) data, ion chromatography (IC) data, scanning electron microscope (SEM) observation data, X-ray diffraction (XRD) data, electron spin resonance (ESR) data, X-ray CT image data, and air permeability measurement data of the materials constituting the storage battery during a life evaluation test. Material analysis parameters obtained by mathematically processing the material analysis data can then be used as explanatory variables for the trained model I. Because obtaining the material analysis data requires disassembling the storage battery, when conducting a life evaluation test, one or more storage batteries are prepared for obtaining test condition data and charge / discharge data, and one or more storage batteries of the same type are prepared for obtaining the material analysis data. Methods for obtaining material analysis parameters from the material analysis data include the following processing steps (1) to (3).

[0163] Processing process (1) is a method for acquiring material analysis parameters from spectra such as XRD, NMR, ESR, and IC. First, a matrix Y is obtained with 1 row and E columns (E is the number of elements, where E is ≥ 2) in which the spectral data in a new state and the spectral data at the time of analysis are arranged as elements in the column direction. The difference between the spectral data in a new state and the spectral data at the time of analysis may be used. The number of elements E employed is approximately 10 to 1000, corresponding to the number of divisions E in the horizontal direction of the spectral data before matrix decomposition. The number of rows in matrix Y is 1 because the evaluation prediction process is performed for each storage battery (for each storage battery). Next, after applying mathematical constraints, matrix Y is decomposed into Y≈V×I (where matrix V is a coefficient matrix with 1 row and F columns, and matrix I is a base matrix with F rows and E columns). The mathematical constraints are set so that matrices V and I are uniquely determined. F is, for example, equal to or less than the number of elements E. As a matrix decomposition method, principal component analysis, non-negative matrix factorization, etc. can be used. For example, when principal component analysis is performed, F coefficients in matrix V (referred to as the "first principal component" to the "Cth principal component" in order from the first column) can be obtained as material analysis parameters. Then, one or more explanatory variables can be selected from the obtained material analysis parameters.

[0164] Processing step (2) is another method for acquiring material analysis parameters from spectra such as XRD, NMR, ESR, and IC. First, parameters characteristic of the curve outline, such as the intensity, integral value, and half-width of the peaks that characterize the material being analyzed, are calculated from the spectral data in the new state and at the time of analysis. Then, the difference between the parameters in the new state and the parameters at the time of analysis is calculated, and the obtained difference can be acquired as material analysis parameters. Then, one or more explanatory variables can be selected from the acquired material analysis parameters.

[0165] Processing (3) is a method of acquiring material analysis parameters from image data such as SEM images. First, a filter process such as a Gaussian filter is applied to the image data in the new state and at the time of analysis to remove noise. Next, a binarization process such as Otsu's binarization is applied to calculate characteristic parameters such as the area ratio of bright and dark areas and circularity. Then, for example, the difference between the parameters in the new state and the parameters at the time of analysis is calculated, and the obtained difference can be acquired as material analysis parameters. Then, one or more explanatory variables can be selected from the acquired material analysis parameters.

[0166] The trained model I is trained to output a predictive evaluation result including predicted battery degradation information that would be obtained for a degraded battery when the multiple explanatory variables are input. More specifically, a database used for training the trained model I is first prepared. The database includes, as training data, charge / discharge characteristics (charge / discharge data) and electrochemical analysis data for storage batteries at the end of their life obtained by conducting a predetermined life evaluation test on multiple storage batteries, and material analysis data for the storage batteries obtained by analyzing the materials of the storage batteries at the end of their life. The dataset further includes test condition data for the life evaluation test, charge / discharge data obtained by conducting an evaluation charge / discharge protocol on multiple storage batteries in various states of degradation, and material analysis data for the storage batteries obtained by disassembling and analyzing multiple storage batteries in various states of degradation. The database also includes multiple explanatory variables extracted from the test condition data, charge / discharge data, and material analysis data, and the multiple explanatory variables are linked to the training data. Next, the trained model I can be generated by using this database to train a machine learning model. The trained model I can be generated by pre-training a machine learning model using supervised learning such as ridge regression, support vector regression, or neural network.

[0167] <Step of acquiring an association model> In the step (S502) of acquiring an association model, an association model is acquired in which the causal relationships between a lifespan determining event that is the basis for determining the lifespan of the storage battery, a plurality of deterioration events, and a plurality of operable factor candidates are previously associated. The association model can be constructed in advance based on correlations between events that are well known to those skilled in the art, empirical rules obtained through experiments, etc. As will be described later, the association model is stored in a memory unit in the operable factor prediction device or a storage device in the operable factor prediction system, and can be acquired from the memory unit or storage device.

[0168] The term "life-determining event" refers to a degradation event that serves as a criterion for determining the degree of degradation of a storage battery and whether it has reached its end of life. The life-determining event is preferably a degradation event used in a general life evaluation test, such as a degradation of the capacity of the storage battery. However, the life-determining event is not limited to the degradation of the capacity of the storage battery, and can be selected from multiple degradation events described below. For example, degradation of the charge / discharge characteristics other than the capacity of the battery, such as a degradation of charge / discharge efficiency, a degradation of performance, or a degradation of safety, can be used as the life-determining event.

[0169] "Degradation events" refer to changes that occur in a storage battery due to repeated charging and discharging of the storage battery, compared to a new storage battery. Examples of degradation events include charge / discharge characteristic degradation events, electrochemical analysis degradation events, and material analysis degradation events. Charge / discharge characteristic degradation events are degradations in characteristics obtained from the charge / discharge data of the storage battery, such as capacity degradation, load degradation (increase in resistance component), no-load degradation (0.2 C capacity degradation), energy density reduction, and power density reduction. Electrochemical analysis degradation events are degradations in characteristics obtained by electrochemical analysis of the storage battery, such as an increase in ohmic resistance, an increase in reaction resistance, an increase in diffusion resistance, a decrease in positive electrode capacity, a decrease in negative electrode capacity, and a deviation in the positive and negative electrode utilization ranges (a deviation in the positive and negative electrode capacity balance due to Li deactivation). Deterioration phenomena determined by material analysis are degradation information obtained by disassembling a storage battery and analyzing the materials, and include, for example, negative electrode SEI formation, increase in current collection resistance, separator clogging, positive electrode-electrolyte interfacial film, negative electrode-electrolyte interfacial film, change in electrolyte composition, change in positive electrode crystal structure, formation of a high-resistance film in the positive electrode, change in negative electrode crystal structure, loss of negative electrode conductivity, formation of a high-resistance film in the negative electrode, and SEI film growth. These degradation phenomena can be quantified using commonly known parameters.

[0170] "Candidate operable factors" refer to design factors that can be changed or manipulated in the manufacture of a storage battery. Examples of candidate operable factors include the density, thickness, volume resistivity, interface resistance, peel strength, porosity and basis weight of the negative electrode active material layer, the type of negative electrode active material, the particle structure, particle diameter, specific surface area, crystal structure, crystallinity, the amount and content of impurities of the negative electrode active material, the type of negative electrode binder and its content, silicon oxide (SiOx) content, the density, thickness, volume resistivity, interface resistance, peel strength, porosity and basis weight of the positive electrode active material layer, the type of positive electrode active material, the particle structure, particle diameter, specific surface area, crystal structure, crystallinity, the amount and content of impurities of the positive electrode active material, the type of positive electrode binder and its content, the type of conductive additive and its content, and the type of electrolyte salt in the electrolyte (e.g., LiPF ６ , LiFSI) and their contents, the type of solvent in the electrolyte and its content, the type of additive in the electrolyte (e.g., vinylene carbonate) and its content, the thickness, pore size, porosity, tortuosity and strength of the separator, the thickness of the coating layer of the separator, the type of filler in the coating layer, the particle size and content of the filler in the coating layer, the type and content of the binder in the coating layer, and the coating position of the coating layer. Examples of the "coating position of the coating layer" include single-sided coating, in which the coating layer is provided only on the surface facing the positive electrode, single-sided coating, in which the coating layer is provided only on the surface facing the negative electrode, and double-sided coating, in which the coating layer is provided on both the positive electrode and negative electrode sides. These candidate manipulable factors are preferred because they may efficiently contribute to improving the lifespan, for example, when capacity degradation is the lifespan-determining event and negative electrode SEI formation is the root cause.

[0171] FIG. 16 is an example of a tree-shaped correlation model based on fault tree analysis (FTA) with a life-determining event as the top event. As illustrated in FIG. 16 , the correlation model is a tree-shaped correlation model based on fault tree analysis (FTA) with a life-determining event as the top event, and can be configured with multiple degradation events at the top level and multiple operable factor candidates at the bottom level. Events connected by solid lines indicate a causal relationship between each other. By using such an FTA-based correlation model, it is possible to predict a wide range of degradation events and operable factors that are causally related with them with minimal analysis, thereby contributing significantly to the efficient improvement of the life of storage batteries. The events and causal relationships illustrated in FIG. 16 are merely examples, and the correlation model is not limited to this example.

[0172] As illustrated in FIG. 16 , in the FTA-based correlation model, the multiple degradation events preferably include, from the highest level down, one or more charge / discharge characteristic degradation events, multiple electrochemical analytical degradation events, and multiple material analytical degradation events. The multiple candidate operable factors can be located at a lower level than each of the multiple material analytical degradation events. Generally, such a hierarchical structure of the FTA-based correlation model for a storage battery allows for prediction of operable factors causally related to a wide range of degradation events with minimal analysis, and improves the readability of the prediction results. In the present disclosure, the lowest-level degradation event directly causally related to the candidate operable factors is also referred to as the “root cause.”

[0173] <Step of calculating the deterioration level> In the step of calculating the deterioration level (S503), multiple deterioration levels representing the degree of deterioration are calculated from the acquired predicted deterioration information, each corresponding to a life-determining event and multiple deterioration events directly or indirectly causally related to the life-determining event. Here, the "deterioration level" is a value indicating the degree to which the life-determining event and multiple deterioration events (charge / discharge characteristics, electrochemical analytical characteristics, and material analytical characteristics) have deteriorated after the deterioration of the storage battery, compared to a new storage battery. The deterioration level is calculated for each life-determining event and multiple deterioration events directly or indirectly causally related to the life-determining event. Even if the deterioration level of a life-determining event is the same value, it may take different values ​​depending on the deterioration event.

[0174] The degree of deterioration of charge / discharge characteristic deterioration phenomena, for example, capacity deterioration, load deterioration (increase in resistive component), no-load deterioration (decrease in non-resistive component), decrease in energy density, and decrease in power density, can be expressed as a percentage of the expected rate of change (decrease rate or increase rate) of each characteristic upon deterioration (e.g., upon reaching the end of its life), with the new state as the reference (0%).

[0175] The degree of deterioration of electrochemical analytical deterioration phenomena, for example, an increase in ohmic resistance, an increase in reaction resistance, an increase in diffusion resistance, a decrease in positive electrode capacity, a decrease in negative electrode capacity, and a deviation in the positive and negative electrode utilization ranges (a deviation in the balance of positive and negative electrode capacities due to Li deactivation), can be expressed as a percentage by taking the new state as the reference (0%) and expressing the rate of change (decrease rate, increase rate, magnitude of deviation, etc.) of each characteristic predicted upon deterioration (for example, at the end of the lifespan).

[0176] The degree of deterioration of material analysis deterioration events can be expressed as a percentage by quantifying any characteristic of the materials of the storage battery, such as the positive electrode active material, the negative electrode active material, the separator material, and the electrolyte material (solvent, salt, and additive), and using the new state as the reference (0%) as the percentage of change (decrease rate, increase rate, etc.) in each characteristic expected upon deterioration (e.g., at the end of life). Examples of material analysis deterioration events to be quantified include positive electrode SEI formation, negative electrode SEI formation, electrolyte starvation, increased current collection resistance, separator clogging, electrolyte composition change, positive electrode crystal structure change, positive electrode conductivity loss, formation of a high-resistance film at the positive electrode-electrolyte interface, cracking of positive electrode active material particles, negative electrode crystal structure change, negative electrode conductivity loss, formation of a high-resistance film at the negative electrode-electrolyte interface, cracking of negative electrode active material particles, and Li deposition. More specifically, for example, the rate of change in the crystallinity of the positive electrode crystals can be expressed as a percentage with the new state as the reference (0%), the rate of increase in the resistance of the negative electrode-electrolyte interface film can be expressed as a percentage with the new state as the reference (0%), the degree of blockage of the separator pores (the rate of reduction in the number of pores) can be expressed as a percentage with the new state as the reference (0%), and the rate of decrease in the separator strength can be expressed as a percentage with the new state as the reference (0%).

[0177] FIG. 17 shows an example of calculating, from predicted degradation information, a life-determining event and the degradation levels corresponding to a plurality of degradation events directly or indirectly causally related to the life-determining event. FIG. 17 shows only a small portion of an exemplary correlation model. In this example, a correlation model based on fault tree analysis (FTA) is used, with capacity degradation as the life-determining event. Potential operational factors are indicated by dark gray shading, and degradation events that are the subject of degradation level calculation are indicated by light gray shading. For example, the degradation level related to capacity degradation is calculated based on the predicted post-degradation capacity (charge / discharge characteristics). The degradation levels related to anode reaction resistance increase, anode diffusion resistance increase, and load degradation are calculated based on predicted discharge curve analysis and GITT (electrochemical analysis data) as well as SEM, NMR, IC, XRD, and ESR data (materials analysis data). Finally, the degradation level related to anode SEI growth is calculated based on predicted NMR and IC data (materials analysis data).

[0178] The degree of capacity degradation can be calculated, for example, from the discharge capacity of a new storage battery (e.g., the cumulative current capacity when discharging from 100% SOC to 0%) and the predicted discharge capacity of the storage battery when deteriorated, using the following formula: degree of capacity degradation = (discharge capacity of storage battery when deteriorated / discharge capacity of storage battery in new state) × 100. This calculation method is an example, and an average of predicted values ​​obtained using multiple calculation formulas may also be obtained. The predicted range of the degree of degradation can be calculated from the difference between the prediction results obtained using the multiple calculation formulas. Figure 17 illustrates an example where the degree of capacity degradation is calculated to be 75% on average.

[0179] The degree of deterioration of the negative electrode SEI growth can be calculated, for example, by obtaining the intensity, area, and half-width of the peak estimated to be derived from the organic SEI component as predicted NMR data, and also obtaining the intensity, area, and half-width of the peak estimated to be derived from the inorganic SEI component as predicted IC data. As an example of the calculation method, each element may be simply linearly combined, or each element may be weighted. The influence of measurement noise can be suppressed by applying some function (such as a step function) to the value of each element. Furthermore, the average of predicted values ​​obtained by multiple calculation formulas may be obtained. The predicted range of the deterioration value can be calculated from the difference in the prediction results obtained by multiple calculation formulas.

[0180] The degree of load degradation and no-load degradation can be calculated by acquiring parameters of a predicted discharge curve analysis. Furthermore, the degree of load degradation and no-load degradation can also be calculated by acquiring and analyzing a DC internal resistance-discharge capacity curve and an open circuit potential-discharge capacity curve predicted by GITT measurement. For example, the negative electrode reaction resistance and negative electrode diffusion resistance can each be calculated based on the ratio of the reaction resistance and diffusion resistance during the negative electrode overvoltage in the predicted voltage relaxation waveform (hereinafter, "voltage relaxation") during GITT measurement. The positive electrode reaction resistance and positive electrode diffusion resistance can also be similarly calculated from the voltage relaxation waveform during GITT measurement. These calculation methods are merely examples, and the negative electrode reaction resistance and negative electrode diffusion resistance, as well as the positive / negative electrode capacity imbalance, may be calculated using multiple calculation formulas. Then, the respective degradation values ​​can be calculated from the negative electrode reaction resistance, negative electrode diffusion resistance, and positive / negative electrode capacity imbalance calculated using multiple methods. As an example of the calculation method, each element may be simply linearly combined, or each element may be weighted. By applying some kind of function (such as a step function) to the value of each element, the effects of measurement noise can be reduced.The predicted range of degradation values ​​can be calculated from the difference in prediction results using multiple calculation formulas.

[0181] Instead of using a value calculated from a single piece of predicted deterioration information as the deterioration value, calculating the deterioration value based on multiple pieces of predicted deterioration information allows for more accurate calculation of the deterioration value. For example, the deterioration value is calculated by arbitrarily combining charge / discharge characteristics, electrochemical analysis data, and material analysis data. More specifically, for example, SEM, NMR, IC, XRD, and ESR have high accuracy in measuring capacity deterioration and resistance deterioration, but are limited to local information. Therefore, by combining electrochemical analysis data that can obtain overall information about the storage battery, such as discharge curve analysis and GITT waveform analysis, and calculating the deterioration level, the deterioration levels of the negative electrode reaction resistance, negative electrode diffusion resistance, and positive / negative electrode capacity imbalance can be accurately determined.

[0182] FIG. 17 illustrates an example of calculating the degree of degradation of each of the negative electrode SEI growth, which is one of the fundamental causes of capacity degradation. Using a similar procedure, for example, as shown in FIG. 18, the degree of degradation can be calculated for degradation events that are fundamental causes other than negative electrode SEI growth, as well as degradation events that are directly or indirectly causally related thereto. FIG. 18 shows only a small portion of an exemplary correlation model. It is preferable to prepare a correlation model that includes degradation events related to the main components of the battery, such as the positive electrode, negative electrode, electrolyte, and separator, and calculate the degree of degradation for each. This allows for comprehensive consideration of a wide range of degradation events in the battery.

[0183] <Step of Simplifying the Association Model> In the step of simplifying the association model (S504), one or more deterioration events whose deterioration level is lower than a threshold are eliminated from the acquired association model, and as a result, any deterioration events and candidate operable factors whose causal relationship with the lifespan determining event has been lost are eliminated, thereby simplifying the association model. A small deterioration level suggests that the range of deterioration level that can be improved (room for improvement) for that deterioration event is small, so by excluding these, it is possible to grasp the fundamental events of deterioration and easily extract candidate operable factors that affect them.

[0184] FIG. 19 shows an example of the process for simplifying the correlation model. In FIG. 19, the portions excluded by the simplification are indicated by dashed lines, and the remaining correlation model is indicated by solid lines. In this example, the threshold for the deterioration level is set to 30%, and deterioration events with a deterioration level of less than 30%, such as positive and negative electrode capacity imbalance, positive electrode capacity decrease, ohmic resistance increase, reaction maldistribution, and separator clogging, are excluded from the correlation model. Furthermore, as a result, the deterioration event of positive electrode crystalline structure collapse and the operable factors causally related to the deleted deterioration events are similarly excluded because their causal relationship with the life-determining events is lost. A single threshold may be set for all deterioration events, or different thresholds may be set for each deterioration event.

[0185] <Step of acquiring impact degree> In the step of acquiring impact degree (S505), when the deterioration degree of the lifespan determining event and the deterioration degrees of the multiple deterioration events are input, an impact degree corresponding to each of one or more of the multiple deterioration events is acquired by inputting the deterioration degree of the lifespan determining event and the deterioration degrees of the multiple deterioration events into a trained model (hereinafter referred to as the "trained model II") that has been trained to output an impact degree that represents the improvement effect on the deterioration degree of the lifespan determining event when one or more of the multiple deterioration events are improved. As will be described later, the trained model II is stored in a memory unit in the operable factor prediction device or a storage device in the operable factor prediction system, and can be acquired from the memory unit or storage device.

[0186] FIG. 20 shows an example of a process for acquiring influence levels. As illustrated in FIG. 20 , by inputting the corresponding deterioration levels for the life-determining event and multiple deterioration events of the simplified correlation model obtained in the previous steps into the trained model II, the influence levels of multiple deterioration events can be acquired. The deterioration level is an index representing the contribution to the life-determining effect, i.e., the extent to which the deterioration level of the life-determining event can be improved when the deterioration level of that deterioration event is improved. While the deterioration value is a value related to that deterioration event alone, the influence level is a relative value that takes into account other deterioration events. If all deterioration events are fully represented within the correlation model, the sum of all the influence levels of the deterioration events in the lower layers should equal the influence level of the upper layers. In other words, the sum of all the influence levels of the lowest layer should equal 100%. However, in reality, there is a limit to the number of deterioration events that can be represented by the correlation model. Therefore, even if all the influence levels of the lowest layer are added together, the sum may be less than 100%, which is acceptable as long as the desired prediction accuracy is ensured.

[0187] The trained model II is trained to output the influence of each degradation event on the degradation level of a life-determining event when the corresponding degradation levels for the life-determining event and multiple degradation events are input. More specifically, a database containing data on the degradation levels of various degradation events, including the life-determining event, measured for multiple storage batteries in various degradation states is first prepared as training data. The database includes data calculated to show the influence of the life-determining event and other degradation events on the degradation level when the degradation level of a certain degradation event is changed. For example, a database is prepared that stores at least one piece of degradation information selected from the group consisting of measured charge / discharge characteristics, electrochemical analysis data, and material analysis data after degradation of the storage battery, and this data is converted into a degradation level using the method described in the "process for calculating the degradation level" above. If necessary, a data structure is constructed in which the degradation levels of the charge / discharge characteristic degradation event, the degradation level of the electrochemical analysis degradation event, the degradation level of the material analysis degradation event, and the values ​​of the operational factors at that time are hierarchically arranged in this order. When specifically associating the degradation level of each degradation event with the degradation level of a lower-level degradation event, the correspondence can be based on correlations between events well known to those skilled in the art, empirical rules obtained through experiments, or the like, as described above in the "Process for Obtaining an Association Model." Next, this database can be used to train a machine learning model to generate a trained model II. The trained model II can be generated by pre-training the machine learning model using supervised learning, such as ridge regression, support vector regression, or neural network. For example, during training, the degree of influence between the degradation level of a certain degradation event and the degradation level of another lower-level degradation event can be learned. Specifically, for example, by performing machine learning on data on the degradation levels of charge / discharge characteristic degradation events and electrochemical analytical degradation events across multiple storage batteries, the degree of influence of the degradation level of a lower-level degradation event on the degradation level of a higher-level degradation event can be calculated, making it possible to determine which degradation level has the greatest influence.

[0188] <Step of outputting operable factors> In the step of outputting operable factors (S506), one or more candidate operable factors that have a direct or indirect causal relationship with the life-determining event in the simplified correlation model are output as operable factors that affect the life of the storage battery.

[0189] In the example shown in FIG. 20, it is found that the root causes of capacity deterioration, which is a life-determining event, are anode SEI growth and cathode crystalline structure collapse. In addition, the manipulable factors that can improve anode SEI growth are anode density, anode porosity, binder type, anode basis weight, SiOx content, anode active material type, and LiPF ６ It can be seen that there are the concentration, the liquid additive A concentration, and the liquid additive B concentration, and that the positive electrode active material is a candidate operable factor that can improve the collapse of the positive electrode crystal structure. Therefore, these candidate operable factors are output as operable factors that affect the life of the storage battery. All of the output operable factors are factors that can improve degradation events that have a large room for improvement, and therefore contribute to efficiently improving the life of the storage battery.

[0190] When the method includes the step (S505) of acquiring the impact degree described above, one or more candidate operable factors that have a direct or indirect causal relationship with the life-determining event in the simplified correlation model and have a direct or indirect causal relationship with one or more deterioration events whose impact degree is equal to or greater than a threshold value can be output as operable factors that affect the life of the storage battery.

[0191] In the example shown in Fig. 21, the threshold of the influence degree is set to 30%, and the causal relationship regarding the collapse of the positive electrode crystalline structure, which is a degradation event with an influence degree of less than 30%, is excluded. As a result, the following factors are considered as operable factors that can improve degradation events with a large margin for improvement and a large effect on improving capacity degradation: negative electrode density, negative electrode porosity, binder type, negative electrode basis weight, SiOx content, negative electrode active material type, LiPF ６ The concentration, liquid additive A concentration, and liquid additive B concentration are output.

[0192] In addition, "outputting" an operable factor includes "sending" the operable factor to another device, and "displaying" the operable factor in any form, such as a graphical user interface (GUI), on a display unit (not shown) of the device itself or another device.

[0193] <Step of predicting change in lifespan of storage battery> The method for predicting an operable factor may further include a step of predicting a change in lifespan of the storage battery. In the step of predicting a change in lifespan of the storage battery, a change in lifespan of the storage battery when each of the output operable factors is changed is predicted, and the prediction result is output.

[0194] Examples of methods for predicting changes in the lifespan of a storage battery include: (i) a method in which, in a simplified correlation model, a theoretical calculation formula representing the relationship between an operable factor and the deterioration degree and / or impact degree of multiple deterioration events is used to associate the operable factor with the input and / or output of a trained model II, and then a forward analysis and / or reverse analysis of the trained model II from a specific operable factor is performed to predict changes in the lifespan of a storage battery when each of the output operable factors is changed (hereinafter referred to as the ``trained model extension type''); and (ii) a method in which, in a simplified correlation model, a theoretical calculation formula representing the relationship between the output operable factor, the deterioration degree of multiple deterioration events, and the deterioration degree of a lifespan determining event is used to construct a simulation model that simulates changes in the lifespan of a storage battery when each of the output operable factors is changed, and then the simulation model is used to simulate changes in the lifespan of a storage battery when each of the output operable factors is changed (hereinafter referred to as the ``simulation model type'').

[0195] (i) Regarding the trained model extension type: In the trained model extension type, theoretical calculation formulas are used to calculate the relationship between the operable factors and the degradation events, and these theoretical calculation formulas are used to extend the input and / or output of the trained model II to the operable factors. Preferably, the input (degradation value) is extended to the operable factors, particularly the degradation value of the degradation event that is the root cause. Using the theoretical calculation formula, a mathematical model that holds true between the operable factors and the degradation value of the degradation event (particularly the root cause) can be derived, predicting how the degradation value of the degradation event will change when the operable factors are changed, and using the changed degradation value as input, predicting the output (impact) due to the change in the degradation value. This allows the operable factors to be indirectly linked to the impact. Then, by reverse-analyzing the trained model II, it is possible to predict the improvement in the lifespan of the storage battery when the operable factors are changed. The reason for expanding the input (deterioration value) of the trained model II into an operable factor in this way is that the output (impact) is a relative value that involves other deterioration factors, while the input (deterioration value) is a value for that deterioration event alone. In other words, by lowering the scope for deriving the theoretical calculation formula to the component or material level rather than the entire storage battery, the technical hurdle for converting the impact on the storage battery's lifespan into an operable factor can be lowered.

[0196] (ii) Simulation Model Type: In the simulation model type, a theoretical calculation formula expressing the relationship between the operable factors, the deterioration levels of multiple deterioration events, and the deterioration level of the lifespan-determining events is used to construct a model that simulates the deterioration levels of the lifespan-determining events when each of the operable factors is changed. To improve the accuracy of the simulation, it is important to match various simulation parameters to the actual behavior of the storage battery. For this matching, the deterioration values ​​calculated from the predicted deterioration information can be used as inputs. In other words, the accuracy of the simulation can be improved by adjusting the simulation parameters so that the deterioration values ​​of each deterioration event become predetermined values ​​that match the behavior of the storage battery. Here, the trained model II can be used as an aid in adjusting the simulation parameters. In other words, to efficiently perform the above-mentioned matching, it is important to narrow down the parameters to be changed (matched) from the countless parameters to a certain extent. By using the calculated deterioration values ​​and the influence level, which is the output of the trained model II, to narrow down the simulation parameters, it is possible to match the parameters to parameters associated with deterioration events with high deterioration values ​​and high influence levels. This makes it possible to improve the efficiency of constructing a simulation model and also to increase the accuracy.

[0197] In addition, "outputting" the predicted change in the lifespan of the storage battery includes "transmitting" the information to another device and "displaying" the information in any form, such as a graphical user interface (GUI), on a display unit (not shown) of the device itself or another device.

[0198] 22 is a diagram showing an example of an operable factor prediction device 800 of the present disclosure. The operable factor prediction device 800 includes a predicted deterioration information acquisition unit 801 that acquires at least one piece of predicted deterioration information selected from the group consisting of predicted charge / discharge characteristics, predicted electrochemical analysis data, and predicted material analysis data after deterioration of a storage battery; a deterioration degree calculation unit 802 that calculates, from the predicted deterioration information, a life-determining event that is a criterion for determining the life of the storage battery, and multiple deterioration degrees that represent the degree of deterioration corresponding to multiple deterioration events that are directly or indirectly causally related to the life-determining event; and a life-determining event, multiple deterioration events, and multiple operable factor candidates. The operable factor prediction device 800 includes: a memory unit 803 that stores an association model in which causal relationships between the lifespan determining events and the lifespan determining events are pre-associated; a simplified association model generation unit 804 that simplifies the association model by eliminating one or more deterioration events whose deterioration level is lower than a threshold value from the association model and, as a result, eliminating any deterioration events and operable factor candidates whose causal relationship with the lifespan determining events has been severed; and an output unit 806 that outputs one or more operable factor candidates that have a direct or indirect causal relationship with the lifespan determining events in the simplified association model as operable factors affecting the lifespan of the storage battery. The functional units, such as the predicted deterioration information acquisition unit, the deterioration level calculation unit, and the memory unit, are connected to each other via a bus and / or an interface unit. The operable factor prediction device 800 may be, for example, an information processing device such as a personal computer, a notebook PC, or a tablet PC. Each functional unit is implemented by a processing unit of the information processing device, such as a CPU (Central Processing Unit), that operates based on a program including an operable factor prediction program pre-stored in the memory unit of the information processing device. The processing unit may be a digital signal processor (DSP), a large scale integration (LSI), an application specific integrated circuit (ASIC), a field-programmable gate array (FPGA), etc. As a result, the manipulable factor prediction device 800 is configured to be able to implement the manipulable factor prediction method of the present disclosure described above.

[0199] 23 is a diagram showing a modified example of the manipulable factor prediction device 800 of the present disclosure. The manipulable factor prediction device 800 preferably further includes an influence acquisition unit 805 that acquires influences corresponding to one or more of the multiple deterioration events by inputting the deterioration level of the lifespan determining event and the deterioration levels of the multiple deterioration events into a trained model II that is trained to output an influence level representing an improvement effect on the deterioration level of the lifespan determining event when one or more of the multiple deterioration events are improved, when the deterioration level of the lifespan determining event and the deterioration levels of the multiple deterioration events are input. In this aspect, the memory unit of the manipulable factor prediction device 800 further stores the trained model II, and the output unit is configured to output, as manipulable factors affecting the lifespan of the storage battery, one or more candidate manipulable factors that have a direct or indirect causal relationship with one or more deterioration events whose influence level is equal to or greater than a threshold in the simplified association model.

[0200] <Predicted Deterioration Information Acquisition Unit> The predicted deterioration information acquisition unit 801 may be a receiving unit that acquires predicted deterioration information from an external device of the manipulable factor prediction device 800. The receiving unit as the predicted deterioration information acquisition unit 801 has a wired communication interface circuit that complies with a communication protocol such as TCP / IP (Transmission Control Protocol / Internet Protocol), and can be communicatively connected to a network in accordance with a communication standard such as Ethernet (registered trademark), or has an antenna that transmits and receives wireless signals and a wireless communication interface circuit that complies with a communication protocol such as wireless LAN, and can be communicatively connected to a network in accordance with a communication standard such as wireless LAN.

[0201] When the output of predicted deterioration information by the trained model I is performed by the processing unit of the operable factor prediction device 800, the operable factor prediction device 800 acquires (receives) test condition data for a life evaluation test, charge / discharge data, and material analysis data from the storage battery charge / discharge device terminal via a network, etc. Then, the processing unit of the operable factor prediction device 800 extracts multiple explanatory variables from this data, inputs the extracted multiple explanatory variables to the trained model I, and outputs predicted deterioration information, and the processing unit serving as the predicted deterioration information acquisition unit 801 acquires the predicted deterioration information.

[0202] <Deterioration Level Calculation Unit> The deterioration level calculation unit 802 may be a processing unit of the manipulable factor prediction device 800. The processing unit as the deterioration level calculation unit 802 calculates, from the predicted deterioration information, a lifespan determining event that is the basis for determining the lifespan of the storage battery, and a plurality of deterioration levels that represent the degree of deterioration corresponding to each of a plurality of deterioration events that are directly or indirectly causally related to the lifespan determining event. Details of the calculation process are as described above.

[0203] <Storage Unit> The storage unit 803 includes a memory device such as a RAM (Random Access Memory) or a ROM (Read Only Memory), a fixed disk device such as a hard disk, or a portable storage device such as a flexible disk or an optical disk. The computer program may be installed into the storage unit using a known setup program from a computer-readable portable recording medium. The portable recording medium is, for example, a CD-ROM (Compact Disc Read Only Memory) or a DVD-ROM (Digital Versatile Disc Read Only Memory). The storage unit 803 stores the manipulable factor prediction program of the present disclosure as well as programs necessary for the operation of the manipulable factor prediction device 800. When the predicted degradation information based on the trained model I is output by the processing unit of the manipulable factor prediction device 800, the storage unit further stores the trained model I. When the processing unit of the manipulable factor prediction device 800 acquires the influence level, the storage unit further stores the trained model II. Details of the trained models I and II are as described above.

[0204] <Simplified Association Model Generator> The simplified association model generator 804 may be a processing unit of the manipulable factor prediction device 800. The processing unit as the simplified association model generator 804 simplifies the association model by excluding, from the association model, one or more deterioration events whose degree of deterioration is lower than a threshold, and as a result, excluding any deterioration events and operable factor candidates whose causal relationship with the lifespan determination event has been lost. Details of the simplification process are as described above.

[0205] <Impact Acquisition Unit> The impact acquisition unit 805 may be a processing unit of the manipulable factor prediction device 800. The processing unit as the impact acquisition unit 805 acquires impacts corresponding to one or more of the multiple degradation events by inputting the deterioration level of the lifespan determining event and the deterioration levels of the multiple degradation events into a trained model II that has been trained to output an impact representing an improvement effect on the deterioration level of the lifespan determining event when one or more of the multiple degradation events are improved, when the deterioration level of the lifespan determining event and the deterioration levels of the multiple degradation events are input. Details of the impact acquisition process are as described above.

[0206] <Output Unit> The output unit 806 may be a transmitter that transmits prediction results, including information on operable factors affecting the battery's lifespan and a prediction of changes in the battery's lifespan when the operable factors are changed, to an external device of the operable factor prediction device 800. The transmitter as the output unit 806 has a wired communication interface circuit conforming to a communication protocol such as TCP / IP (Transmission Control Protocol / Internet Protocol) and can communicate with a network according to a communication standard such as Ethernet (registered trademark), or has an antenna for transmitting and receiving wireless signals and a wireless communication interface circuit conforming to a communication protocol such as wireless LAN and can communicate with a network according to a communication standard such as wireless LAN. The prediction results transmitted to the external device can be displayed in any form, such as a graphical user interface (GUI), on the display unit of the external device. Examples of the external device include a business operator terminal that operates the operable factor prediction device of the present disclosure, or a customer terminal that uses the operable factor prediction device of the present disclosure.

[0207] The output unit 806 may be a display unit of the manipulable factor prediction device 800. In this case, the display unit can display (output) the prediction result to the user of the manipulable factor prediction device 800 in any form such as a graphical user interface (GUI).

[0208] 《Manipulable Factor Prediction System》 The operable factor prediction system disclosed herein includes: an acquisition device that acquires at least one predicted deterioration information selected from the group consisting of predicted charge / discharge characteristics, predicted electrochemical analysis data, and predicted material analysis data after deterioration of a storage battery; a deterioration degree calculation device that calculates, from the predicted deterioration information, a life-determining event that serves as a basis for determining the life of the storage battery, and a plurality of deterioration levels that indicate the degree of deterioration, each corresponding to a plurality of deterioration events that are directly or indirectly causally related to the life-determining event; a storage device that stores an association model that pre-associates causal relationships between the life-determining event, the plurality of deterioration events, and a plurality of candidate operable factors; a simplified association model generation device that simplifies the association model by excluding from the association model one or more deterioration events whose deterioration level is lower than a threshold, and as a result, excluding any deterioration events and candidate operable factors that have lost their causal relationship with the life-determining event; and an output device that outputs one or more candidate operable factors that are directly or indirectly causally related to the life-determining event in the simplified association model as operable factors that affect the life of the storage battery. Each device may be an information processing device such as a personal computer, a notebook PC, or a tablet PC, which are connected to each other so as to be able to communicate with each other via a wired and / or wireless network.

[0209] In the operable factor prediction system, as long as the operable factor prediction method of the present disclosure can be implemented as an entire system connected to each other in a manner that allows communication, any of the devices constituting the system may function as a predicted deterioration information acquisition device, a deterioration degree calculation device, a storage device, a simplified association model generation device, an influence acquisition device, and an output device. Furthermore, a single device may have multiple functions of these devices, or multiple devices may be configured to implement one function. For details on the operation of each device, please refer to the description of each functional unit of the operable factor prediction device described above.

[0210] <<Operable Factor Prediction Program>> The operable factor prediction program of the present disclosure includes having a computer execute the following steps: calculating a life-determining event that serves as a basis for determining the life of a storage battery, and a plurality of deterioration levels that indicate the degree of deterioration, each corresponding to a plurality of deterioration events that are directly or indirectly causally related to the life-determining event, from at least one predicted deterioration information selected from the group consisting of predicted charge / discharge characteristics, predicted electrochemical analysis data, and predicted material analysis data after the storage battery has deteriorated; simplifying the association model that previously associates the causal relationships between the life-determining event, the plurality of deterioration events, and a plurality of operable factor candidates by eliminating one or more deterioration events with a deterioration level lower than a threshold, and as a result, eliminating any deterioration events and operable factor candidates whose causal relationship with the life-determining event has been severed; and outputting the one or more operable factor candidates that are directly or indirectly causally related to the life-determining event in the simplified association model as operable factors that affect the life of the storage battery.

[0211] The operationable factor prediction program is stored in a memory unit or a storage device in the operationable factor prediction device or system, and causes the computer of the operationable factor prediction device or system to operate so as to implement the operationable factor prediction method of the present disclosure. For details of each operation, please refer to the descriptions of the operationable factor prediction method, device, and program described above.

[0212] Combination of Evaluation Result Prediction and Manipulable Factor Prediction The operable factor prediction device 800 of the present disclosure, as illustrated in FIGS. 22 and 23 , may (1) double as the evaluation result prediction device of the present disclosure described above, or (2) be a device physically independent from the evaluation result prediction device. In the case of (1), the evaluation result prediction device further includes the configuration of the operable factor prediction device 800. Such a device can be said to be an evaluation result prediction device with an operable factor prediction function, or an operable factor prediction device with an evaluation result prediction function. For example, the charging / discharging device terminal 200, the business operator server 300, or the shared server 700 serving as the evaluation result prediction device may further include the configuration of the manipulable factor prediction device 800. In the case of (2), another information processing device communicatively connected to the evaluation result prediction device may include the configuration of the manipulable factor prediction device 800. The other information processing device may be the charging / discharging device terminal 200, the business operator server 300, the shared server 700, or an information processing device other than the evaluation result prediction device. In either aspect (1) or (2), the operable factor prediction device 800 can predict operable factors using the predicted evaluation results predicted by the evaluation result prediction device of the present disclosure, thereby making it possible to more efficiently improve the battery life of the storage battery.

[0213] The operable factor prediction system of the present disclosure may also serve as the evaluation result prediction system of the present disclosure. That is, the evaluation result prediction system further includes a configuration as an operable factor prediction system. Such a system can be described as an evaluation result prediction system with an operable factor prediction function, or as an operable factor prediction system with an evaluation result prediction function. For example, the charging / discharging device terminal 200, the business operator server 300, the shared server 700, or other information processing devices communicatively connected to the evaluation result prediction system, which constitute the evaluation result prediction system, may individually or collectively include a configuration as an operable factor prediction system. With this configuration, the operable factor prediction system can predict operable factors using the predicted evaluation results predicted by the evaluation result prediction method of the present disclosure, thereby more efficiently improving the lifespan of the storage battery.

[0214] More specifically, FIG. 24 is a diagram showing a schematic configuration of the evaluation result and operable factor prediction system of the present disclosure. The evaluation result and operable factor prediction system 2 includes a charging / discharging device 100, a charging / discharging device terminal 200 communicatively connected to the charging / discharging device 100, a business operator server 900 communicatively connected to the charging / discharging device terminal 200 via a business operator's private network N1, and a virtual environment 400 constructed on a network N2 and communicatively connected to the business operator server 900. The evaluation result and operable factor prediction system 2 is accessible from terminal devices such as a business operator terminal 500 and a customer terminal 600, which are indicated by dashed lines. The business operator terminal 500 is communicatively connected to the charging / discharging device terminal 200 and the business operator server 900 via the business operator's private network N1, and is communicatively connected to the virtual environment 400 and the customer terminal 600 via the network N2. The customer terminal 600 is communicatively connected to the virtual environment 400 and the business operator terminal 500 via the network N2.

[0215] 24 is an "evaluation result and operable factor prediction device" that can predict the evaluation results of a predetermined life evaluation test and the energy density and power density of a storage battery measured sequentially using a predetermined charge / discharge protocol, and can predict operable factors based on the predicted evaluation results. The business operator server 900 includes an acquisition unit (receiving unit) 901 that can acquire test condition data, charge / discharge data, material analysis data, etc. of the storage battery, an explanatory variable extraction unit 902 that extracts multiple explanatory variables from the test condition data, charge / discharge data, and material analysis data, a storage unit 903 that stores a trained model I that has been trained to output a predicted evaluation result when multiple explanatory variables are input, and a database, etc., an evaluation prediction unit 904 that inputs the multiple explanatory variables to the trained model to obtain a predicted evaluation result and calculates the prediction accuracy of the predicted evaluation result, and an output unit (transmitting unit) 905 that outputs information including the predicted evaluation result and the prediction accuracy.

[0216] 24 further includes a control unit 906 that outputs a command to (i) terminate the evaluation charge / discharge protocol if the prediction accuracy is equal to or greater than a predetermined threshold, and (ii) continue the evaluation charge / discharge protocol if the prediction accuracy is less than the predetermined threshold. The business operator server 900 further includes a learning processing unit 907 that can generate and re-learn the trained model I.

[0217] The acquisition unit (receiving unit) 901 of the business operator server 900 in FIG. 24 also functions as a predicted deterioration information acquisition unit that acquires at least one piece of predicted deterioration information selected from the group consisting of predicted charge / discharge characteristics, predicted electrochemical analysis data, and predicted material analysis data after deterioration of the storage battery. The business operator server 900 further includes a deterioration degree calculation unit 908 that calculates, from the predicted deterioration information, a life-determining event that serves as a basis for determining the life of the storage battery, and multiple deterioration levels that indicate the degree of deterioration corresponding to multiple deterioration events that are directly or indirectly causally related to the life-determining event. The storage unit 903 further stores an association model that pre-associates causal relationships between the life-determining event, multiple deterioration events, and multiple candidate operable factors. The business operator server 900 further includes a simplified association model generation unit 909 that simplifies the association model by excluding from the association model one or more deterioration events whose deterioration level is lower than a threshold and, as a result, excluding any deterioration events and candidate operable factors whose causal relationship with the life-determining event has been severed. The output unit (transmission unit) 905 outputs one or more candidate operable factors that have a direct or indirect causal relationship with the life-determining event in the simplified correlation model as operable factors that affect the life of the storage battery.

[0218] The business operator server 900 preferably further includes an impact acquisition unit (not shown) that acquires impacts corresponding to one or more of the plurality of degradation events by inputting the degradation level of the lifespan determining event and the degradation levels of the plurality of degradation events into a trained model II that has been trained to output, when the degradation level of the lifespan determining event and the degradation levels of the plurality of degradation events are input, an impact level that represents an improvement effect on the degradation level of the lifespan determining event when one or more of the plurality of degradation events are improved. In this aspect, the memory unit 903 of the business operator server 900 further stores the trained model II, and the output unit (transmission unit) 905 is further configured to output, as an operational factor affecting the lifespan of the storage battery, one or more candidate operable factors that have a direct or indirect causal relationship with one or more degradation events whose impact level is equal to or greater than a threshold in the simplified association model.

[0219] The evaluation result and operable factor prediction system 2, having the above configuration, can predict the evaluation results that would be obtained if a predetermined life evaluation test were performed on the storage battery to be evaluated, and the deterioration information of the storage battery at the end of its life, based on the test condition data in the life evaluation test and the charge / discharge data obtained by charging and discharging the storage battery to be evaluated based on the evaluation charge / discharge protocol during the life evaluation test. Furthermore, the evaluation result and operable factor prediction system 2 can predict the evaluation results that would be obtained if a predetermined charge / discharge protocol were performed on the storage battery to be evaluated, based on the charge / discharge data obtained by charging and discharging the storage battery to be evaluated based on the evaluation charge / discharge protocol. Furthermore, the evaluation result and operable factor prediction system 2 can predict operable factors using the predicted evaluation results.

[0220] Examples and comparative examples of the present disclosure will be described below, but the present disclosure is not limited to the following examples and comparative examples. The evaluation result prediction device and system of the present disclosure were used to predict the number of cycles to reach life measured in a predetermined life evaluation test.

[0221]

[0222] In Examples 1 to 4, a prediction model was constructed using the support vector regression method, with the natural logarithm of the number of cycles to reach the end of life of the battery as the objective variable. The test conditions set in the life evaluation test, power density, and energy density were prepared as explanatory variables, and changes in prediction performance were verified by comparing the cases where power density and energy density were added to the test conditions as explanatory variables (○) and where they were not added (×).

[0223] When only the energy density was added to the test conditions as an explanatory variable (Example 3) and when only the power density was added (Example 2), the coefficient of determination (R ２ The coefficient of determination (COD) increased and the mean absolute error (MAE) decreased. Furthermore, when both energy density and power density were added as explanatory variables (Example 4), the coefficient of determination increased and the mean absolute error decreased more than when only one of them was used (Examples 2 and 3). Therefore, it was demonstrated that adding energy density and power density as explanatory variables may improve the accuracy of prediction at the end of life.

[0224] The predicted number of cycles to reach life was the number of cycles until the voltage retention rate or capacity retention rate of the storage battery fell below a predetermined reference value or the storage battery reached a predetermined upper temperature limit when an evaluation charge / discharge protocol was performed during the life evaluation test. The test conditions included conditions such as current, voltage, and temperature of the charge / discharge cycle in the deterioration step of the storage battery life evaluation test. The energy density included the value of the current capacity or power capacity per volume or weight of the storage battery. The power density included the output characteristics at high current values ​​and the difference in the value of the current capacity or power capacity at different current values.

[0225] In Examples 1 to 4, data were used that were obtained by carrying out a life evaluation test on 15 samples of the same type of lithium ion battery under different degradation conditions. In Examples 1 to 4, the prediction accuracy was evaluated using cross-validation. The coefficient of determination (R ２ The mean abso...

Claims

1. An evaluation result prediction device that predicts at least one evaluation result selected from a post-test voltage retention rate, a post-test capacity retention rate, a time to reach life, or a number of cycles to reach life of a storage battery, which are measured in a predetermined life evaluation test, wherein the evaluation result prediction device comprises: an acquisition unit that acquires at least one test condition data selected from a temperature, a current value, a voltage, or a charge / discharge rest time of the storage battery, and charge / discharge data of the storage battery based on an evaluation charge / discharge protocol that includes measuring energy density and power density during the life evaluation test; an explanatory variable extraction unit that extracts a plurality of explanatory variables including one or more test condition parameters selected from the test condition data and electrochemical parameters including the energy density and the power density calculated from the charge / discharge data; a storage unit that stores a trained model that, when the plurality of explanatory variables are input, has been trained to output a predicted evaluation result including the evaluation result measured in the life evaluation test; and an evaluation prediction unit that acquires the predicted evaluation result by inputting the plurality of explanatory variables into the trained model, and calculates the prediction accuracy of the predicted evaluation result. an output unit that outputs information including the predicted evaluation result and the prediction accuracy.

2. The evaluation result prediction device according to claim 1, wherein the predetermined life evaluation test includes: a degradation step in which the storage battery is stored at a predetermined temperature for a fixed period (X) or a fixed number (Y) of charge / discharge cycles are repeated at a predetermined temperature; and an evaluation step including: (i) measuring the voltage retention rate and / or capacity retention rate of the storage battery after completion of the degradation step, or (ii) measuring the life reaching time or the number of life reaching cycles until the voltage retention rate or capacity retention rate of the storage battery falls below a predetermined reference value or until the storage battery reaches a predetermined upper limit temperature.

3. The evaluation charge / discharge protocol includes measuring the energy density and the power density in a single charge / discharge process, and the single charge / discharge process includes measuring the energy density by discharging the storage battery at the predetermined temperature and a predetermined current value, and temporarily increasing the maximum current value I when the charging rate reaches a target charging rate during the discharging process. ｍａｘ and measuring the power density at the target charging rate, and then returning the current value to the predetermined value again, this process being performed each time one or more of the target charging rates is reached, thereby measuring the energy density and the power density in a single discharge process; the trained model is further trained to output a predicted evaluation result including the energy density and power density measured in a predetermined charging / discharging protocol when the plurality of explanatory variables are input; the evaluation prediction unit obtains the predicted evaluation result including the energy density and the power density measured in the predetermined charging / discharging protocol by inputting the plurality of explanatory variables into the trained model, and calculates a prediction accuracy of the predicted evaluation result; and the output unit outputs information including the predicted evaluation result and the prediction accuracy.

4. The predetermined charge / discharge protocol includes an energy density measurement step of discharging the storage battery at a first temperature and a first current value to measure the energy density, and a charge or discharge process that is different from the energy density measurement step in temperature, charge or discharge, or current value, or two or more of these are different, and adjusts the storage battery to a target charge rate and discharges the storage battery to a maximum current value I ｍａｘ and one or more power density measurement steps, wherein the single charge / discharge process comprises measuring an energy density by discharging the storage battery at the first temperature and at a second current value greater than the first current value, and temporarily outputting a maximum current value I when the charging rate reaches the target charging rate during the discharging process. ｍａｘ and measuring the power density at the target charging rate, and then returning to the second current value again, this is performed each time one or more of the target charging rates are reached, thereby measuring the energy density and the power density in a single discharge process; the evaluation charge / discharge protocol includes first to n-th single charge / discharge processes, each of which measures the energy density and the power density at first to n-th different temperatures (n is an integer of 2 or more) in a single charge / discharge process, and the first to n-th single charge / discharge processes are configured in an order that increases the prediction accuracy or completes the temperature adjustment in a shorter time; the evaluation result prediction device is configured to obtain the predicted evaluation result based on charge / discharge data of a first single charge / discharge process and calculate the prediction accuracy; the evaluation result prediction device further includes a control unit, which: (i) terminates the evaluation charge / discharge protocol if the prediction accuracy is equal to or greater than a predetermined threshold; 4. The evaluation result prediction device according to claim 3, wherein the evaluation result prediction device is controlled so that, if the prediction accuracy is less than the threshold, a next single charge / discharge process of the evaluation charge / discharge protocol is performed, further charge / discharge data is acquired, and acquisition of the predicted evaluation result and calculation of the prediction accuracy based on the accumulated charge / discharge data is repeated until the prediction accuracy becomes equal to or greater than the threshold or until an n-th single charge / discharge process is completed.

5. The evaluation result prediction device according to any one of claims 2 to 4, wherein the evaluation charge / discharge protocol includes a plurality of evaluation charge / discharge protocols, each including measuring the energy density and the power density in a single charge / discharge process at regular intervals shorter than X or at regular intervals for a certain number of charge / discharge cycles less than Y during the life evaluation test; and the evaluation result prediction device is configured to obtain the predicted evaluation result based on the test condition data and charge / discharge data of the single charge / discharge process, and calculate the prediction accuracy; and the evaluation result prediction device further includes a control unit, which controls the evaluation result prediction device to: (i) terminate the life evaluation test if the prediction accuracy is equal to or greater than a predetermined threshold; and (ii) continue the life evaluation test and perform a next evaluation charge / discharge protocol to obtain further charge / discharge data if the prediction accuracy is less than the threshold, and repeat the process of obtaining the predicted evaluation result and calculating the prediction accuracy based on the test condition data and accumulated charge / discharge data, until the prediction accuracy becomes equal to or greater than the threshold.

6. The evaluation result prediction device according to any one of claims 1 to 4, wherein the acquisition unit is configured to further acquire at least one material analysis data selected from an X-ray diffraction (XRD) spectrum, a nuclear magnetic resonance (NMR) spectrum, an electron spin resonance (ESR) spectrum, a scanning electron microscope (SEM) image, or an ion chromatography (IC) spectrum of a material constituting the storage battery, and the plurality of explanatory variables further include, in addition to the test condition parameters and the electrochemical parameters, one or more material analysis parameters obtained by mathematically processing the material analysis data.

7. The evaluation result prediction device according to any one of claims 1 to 4, wherein the plurality of explanatory variables further include, in addition to the test condition parameters and the electrochemical parameters, one or more mathematical parameters obtained by mathematically processing the shape of the discharge curve in a single charge / discharge process, and the plurality of explanatory variables are explanatory variables selected in advance by excluding one or more parameters that are multicollinear with other parameters so as to improve the prediction accuracy of the trained model or improve the interpretability of the predicted evaluation results.

8. The evaluation result prediction device according to claim 7, wherein the mathematical parameters are explanatory variables pre-selected from C coefficients in matrix W obtained by decomposing a 1-row, B-column matrix X, in which voltage (V) values corresponding to B (B≧2) different state of charge (SOC) values are arranged as elements in the column direction, for the discharge curve data in the single charge / discharge process, into X≈W×H (where matrix W is a 1-row, C-column coefficient matrix, and matrix H is a C-row, B-column basis matrix) after adding mathematical constraints.

9. The evaluation result prediction device according to any one of claims 1 to 4, wherein the predicted evaluation result further includes predicted deterioration information of the storage battery at the end of its life.

10. The evaluation result prediction device described in any one of claims 1 to 4, wherein the output unit is configured to output information including the predicted evaluation result and the prediction accuracy to a virtual environment accessible by a user terminal, and is further configured to terminate the evaluation charge / discharge protocol when a command to terminate the evaluation charge / discharge protocol is received from the virtual environment.

11. An evaluation result prediction system that predicts at least one evaluation result selected from a post-test voltage retention rate, a post-test capacity retention rate, a time to reach life, or a number of cycles to reach life of a storage battery measured in a predetermined life evaluation test, the evaluation result prediction system comprising: an acquisition device that acquires at least one test condition data selected from a temperature, a current value, a voltage, or a charge / discharge rest time of the storage battery, and charge / discharge data of the storage battery based on an evaluation charge / discharge protocol that includes measuring energy density and power density during the life evaluation test; an explanatory variable extraction device that extracts a plurality of explanatory variables including one or more test condition parameters selected from the test condition data and electrochemical parameters including the energy density and the power density calculated from the charge / discharge data; a storage device that stores a trained model that, when the plurality of explanatory variables are input, is trained to output a predicted evaluation result including the evaluation result measured in the life evaluation test; and an evaluation prediction device that acquires the predicted evaluation result by inputting the plurality of explanatory variables into the trained model, and calculates the prediction accuracy of the predicted evaluation result. an output device that outputs information including the predicted evaluation result and the prediction accuracy.

12. The evaluation result prediction system according to claim 11, wherein the predetermined life evaluation test includes: a degradation step in which the storage battery is stored at a predetermined temperature for a certain period (X) or a certain number (Y) of charge / discharge cycles are repeated at a predetermined temperature; and an evaluation step including: (i) measuring the voltage retention rate and / or capacity retention rate of the storage battery after completion of the degradation step, or (ii) measuring the life reaching time or the number of life reaching cycles until the voltage retention rate or capacity retention rate of the storage battery falls below a predetermined reference value or until the storage battery reaches a predetermined upper limit temperature.

13. The evaluation charge / discharge protocol includes measuring the energy density and the power density in a single charge / discharge process, and the single charge / discharge process includes measuring the energy density by discharging the storage battery at the predetermined temperature and a predetermined current value, and temporarily increasing the maximum current value I when the charging rate reaches a target charging rate during the discharging process. ｍａｘ and measuring the power density at the target charging rate, and then returning the current value to the predetermined value again, each time one or more of the target charging rates is reached, thereby measuring the energy density and the power density in a single discharge process; the trained model is further trained to output a predicted evaluation result including the energy density and power density measured in a predetermined charging / discharging protocol when the plurality of explanatory variables are input; the evaluation prediction device obtains the predicted evaluation result including the energy density and the power density measured in the predetermined charging / discharging protocol by inputting the plurality of explanatory variables into the trained model, and calculates the prediction accuracy of the predicted evaluation result; and the output device outputs information including the predicted evaluation result and the prediction accuracy.

14. A method for predicting at least one evaluation result selected from a post-test voltage retention rate, a post-test capacity retention rate, a time to reach life, or a number of cycles to reach life of a storage battery measured in a predetermined life evaluation test, the method comprising: acquiring at least one test condition data selected from a temperature, a current value, a voltage, or a charge / discharge rest time of the storage battery, and charge / discharge data of the storage battery based on an evaluation charge / discharge protocol including measuring energy density and power density during the life evaluation test; extracting a plurality of explanatory variables including one or more test condition parameters selected from the test condition data and electrochemical parameters including the energy density and the power density calculated from the charge / discharge data; obtaining the predicted evaluation result by inputting the plurality of explanatory variables into a trained model that is trained to output a predicted evaluation result including the evaluation result measured in the life evaluation test when the plurality of explanatory variables are input; calculating the prediction accuracy of the predicted evaluation result; and outputting information including the predicted evaluation result and the prediction accuracy.

15. The evaluation result prediction method according to claim 14, wherein the predetermined life evaluation test includes: a degradation step in which the storage battery is stored at a predetermined temperature for a certain period (X) or a certain number (Y) of charge / discharge cycles are repeated at a predetermined temperature; and an evaluation step including: (i) measuring the voltage retention rate and / or capacity retention rate of the storage battery after completion of the degradation step, or (ii) measuring the life reaching time or the number of life reaching cycles until the voltage retention rate or capacity retention rate of the storage battery falls below a predetermined reference value or until the storage battery reaches a predetermined upper limit temperature.

16. The evaluation charge / discharge protocol includes measuring the energy density and the power density in a single charge / discharge process, and the single charge / discharge process includes measuring the energy density by discharging the storage battery at the predetermined temperature and a predetermined current value, and temporarily increasing the maximum current value I when the charging rate reaches a target charging rate during the discharging process. ｍａｘ and measuring the power density at the target charging rate, and then returning the current value to the predetermined current value again, each time one or more of the target charging rates is reached, thereby measuring the energy density and the power density in a single discharge process; acquiring the predicted evaluation result including the energy density and the power density measured in a predetermined charging / discharging protocol by inputting the plurality of explanatory variables into the trained model, which has been further trained to output a predicted evaluation result including the energy density and the power density measured in the predetermined charging / discharging protocol when the plurality of explanatory variables are input, and calculating a prediction accuracy of the predicted evaluation result; and outputting information including the predicted evaluation result and the prediction accuracy.

17. An evaluation result prediction program that predicts at least one evaluation result selected from a post-test voltage retention rate, a post-test capacity retention rate, a time to reach life, or a number of cycles to reach life of a storage battery, which are measured in a predetermined life evaluation test, wherein the evaluation result prediction program causes a computer to execute the following steps: acquire at least one test condition data selected from a temperature, a current value, a voltage, or a charge / discharge rest time of the storage battery, and charge / discharge data of the storage battery based on an evaluation charge / discharge protocol that includes measuring energy density and power density during the life evaluation test; extract a plurality of explanatory variables including one or more test condition parameters selected from the test condition data and electrochemical parameters including the energy density and the power density calculated from the charge / discharge data; acquire the predicted evaluation result by inputting the plurality of explanatory variables into a trained model that is trained to output a predicted evaluation result including the evaluation result measured in the life evaluation test when the plurality of explanatory variables are input; calculate the prediction accuracy of the predicted evaluation result; and output information including the predicted evaluation result and the prediction accuracy.

18. The evaluation result prediction program according to claim 17, wherein the predetermined life evaluation test includes: a degradation step in which a storage battery is stored at a predetermined temperature for a certain period (X) or a certain number (Y) of charge / discharge cycles are repeated at a predetermined temperature; and an evaluation step including: (i) measuring the voltage retention rate and / or capacity retention rate of the storage battery after completion of the degradation step, or (ii) measuring the life reaching time or the number of life reaching cycles until the voltage retention rate or capacity retention rate of the storage battery falls below a predetermined reference value or until the storage battery reaches a predetermined upper limit temperature.

19. The evaluation charge / discharge protocol includes measuring the energy density and the power density in a single charge / discharge process, and the single charge / discharge process includes measuring the energy density by discharging the storage battery at the predetermined temperature and a predetermined current value, and temporarily increasing the maximum current value I when the charging rate reaches a target charging rate during the discharging process. ｍａｘ and measuring the power density at the target charging rate, and then returning the current value to the predetermined current value again, each time one or more of the target charging rates is reached, thereby measuring the energy density and the power density in a single discharge process; and the evaluation result prediction program causes a computer to execute the following: by inputting the plurality of explanatory variables into the trained model, which has been further trained to output a predicted evaluation result including the energy density and the power density measured in a predetermined charging / discharging protocol when the plurality of explanatory variables are input, obtain the predicted evaluation result including the energy density and the power density measured in the predetermined charging / discharging protocol, calculate a prediction accuracy of the predicted evaluation result, and output information including the predicted evaluation result and the prediction accuracy.

20. A method for predicting operable factors affecting the life of a storage battery, the method comprising: a step of calculating a life-determining event serving as a basis for determining the life of the storage battery and a plurality of deterioration levels representing the degree of deterioration corresponding to a plurality of deterioration events that are directly or indirectly causally related to the life-determining event, from at least one piece of predicted deterioration information selected from the group consisting of predicted charge / discharge characteristics, predicted electrochemical analysis data, and predicted material analysis data after deterioration of the storage battery; a step of simplifying an association model that previously associates the causal relationships between the life-determining event, the plurality of deterioration events, and a plurality of operable factor candidates by excluding one or more deterioration events whose deterioration level is lower than a threshold, and as a result, excluding any deterioration events and operable factor candidates whose causal relationship with the life-determining event has been severed; and a step of outputting one or more operable factor candidates that are directly or indirectly causally related to the life-determining event in the simplified association model as operable factors affecting the life of the storage battery. A method for predicting an operable factor, comprising:

21. The method further includes obtaining the influence corresponding to one or more of the plurality of degradation events by inputting the deterioration degree of the lifespan determining event and the deterioration degrees of the plurality of degradation events into a trained model that is trained to output, when the deterioration degree of the lifespan determining event and the deterioration degrees of the plurality of degradation events are input, an influence degree representing the improvement effect on the deterioration degree of the lifespan determining event when one or more of the plurality of degradation events are improved, and outputting, in the simplified correlation model, one or more candidate operable factors that have a direct or indirect causal relationship with one or more of the degradation events whose influence degree is equal to or greater than a threshold, as operable factors affecting the lifespan of the storage battery.

22. The method for predicting an operable factor according to claim 21, further comprising: using a theoretical calculation formula representing the relationship between the operable factor and the deterioration level and / or the impact level of the multiple deterioration events in the simplified association model to associate the operable factor with the input and / or output of the trained model, and predicting changes in the lifespan of the storage battery when each of the output operable factors is changed by performing forward analysis and / or reverse analysis of the trained model from a specific operable factor, and outputting the prediction result; and using a theoretical calculation formula representing the relationship between the output operable factor, the deterioration level of the multiple deterioration events, and the deterioration level of a lifespan determining event in the simplified association model to construct a simulation model simulating changes in the lifespan of the storage battery when each of the output operable factors is changed, and using the simulation model to simulate changes in the lifespan of the storage battery when each of the output operable factors is changed, and outputting the simulation result.

23. The predicted deterioration information includes predicted charge / discharge characteristics, predicted electrochemical analysis data, and predicted material analysis data of the storage battery at the end of its life, the predicted charge / discharge characteristics including at least one selected from the group consisting of energy density, power density, capacity, and charge / discharge efficiency of the storage battery at the end of its life, the predicted electrochemical analysis data including at least one selected from the group consisting of charge curve analysis (CCA), discharge curve analysis, constant current intermittent titration (GITT), and electrochemical impedance analysis (EIS) of the storage battery at the end of its life, and the predicted material analysis data including at least one selected from the group consisting of nuclear magnetic resonance (NMR) data, ion chromatography (IC) data, scanning electron microscope (SEM) observation data, X-ray diffraction (XRD) data, electron spin resonance (ESR) data, X-ray CT image data, and air permeability measurement data of materials constituting the storage battery at the end of its life, The operable factor candidates include the density, thickness, volume resistivity, interface resistance, peel strength, porosity and basis weight of the negative electrode active material layer, the type of negative electrode active material, the particle structure, particle size, specific surface area, crystal structure, crystallinity, the amount and content of impurities of the negative electrode active material, the type of negative electrode binder and its content, silicon oxide (SiOx) content, the density, thickness, volume resistivity, interface resistance, peel strength, porosity and basis weight of the positive electrode active material layer, the type of positive electrode active material, the particle structure, particle size, specific surface area, crystal structure, crystallinity, the amount and content of impurities of the positive electrode active material, the type of positive electrode binder and its content, and the material of the conductive additive. the type and content thereof, the type of electrolyte salt in the electrolyte solution and content thereof, the type of solvent in the electrolyte solution and content thereof, the type of additive in the electrolyte solution and content thereof, the thickness, pore size, porosity, tortuosity and strength of the separator, the thickness of the coating layer of the separator, the type of filler in the coating layer, the particle size and content of the filler in the coating layer, the type and content of the binder in the coating layer, and the coating position of the coating layer (single-sided coating on the positive electrode side, single-sided coating on the negative electrode side, or double-sided coating).

24. An apparatus for predicting operable factors that affect the life of a storage battery, the apparatus comprising: an acquisition unit that acquires at least one piece of predicted deterioration information selected from the group consisting of predicted charge / discharge characteristics, predicted electrochemical analysis data, and predicted material analysis data after deterioration of the storage battery; a deterioration degree calculation unit that calculates, from the predicted deterioration information, a life determining event that is the basis for determining the life of the storage battery, and multiple deterioration levels that indicate the degree of deterioration, each corresponding to a plurality of deterioration events that are directly or indirectly causally related to the life determining event; a storage unit that stores an association model that previously associates the causal relationships between the life determining event, the multiple deterioration events, and multiple candidate operable factors; and a simplified association model generation unit that simplifies the association model by excluding from the association model one or more deterioration events whose deterioration level is lower than a threshold, and as a result, excluding any deterioration events and candidate operable factors that have lost their causal relationship with the life determining event; an output unit that outputs one or more candidate operable factors that have a direct or indirect causal relationship with the life-determining event in the simplified correlation model as operable factors that affect the life of the storage battery.

25. A system for predicting operable factors that affect the life of a storage battery, comprising: an acquisition device that acquires at least one piece of predicted deterioration information selected from the group consisting of predicted charge / discharge characteristics, predicted electrochemical analysis data, and predicted material analysis data after the storage battery has deteriorated; a deterioration degree calculation device that calculates, from the predicted deterioration information, a life determining event that is the basis for determining the life of the storage battery, and multiple deterioration levels that indicate the degree of deterioration, each corresponding to a plurality of deterioration events that are directly or indirectly causally related to the life determining event; a storage device that stores an association model that previously associates the causal relationships between the life determining event, the multiple deterioration events, and multiple candidate operable factors; and a simplified association model generation device that simplifies the association model by excluding from the association model one or more deterioration events whose deterioration level is lower than a threshold, and as a result, excluding any deterioration events and candidate operable factors that have lost their causal relationship with the life determining event; an output device that outputs one or more candidate operable factors that have a direct or indirect causal relationship with the life-determining event in the simplified correlation model as operable factors that affect the life of the storage battery.

26. A program for predicting operable factors that affect the life of a storage battery, the program comprising the following steps: calculating a life-determining event that is a basis for determining the life of the storage battery and a plurality of deterioration levels that indicate the degree of deterioration, each corresponding to a plurality of deterioration events that are directly or indirectly causally related to the life-determining event, from at least one piece of predicted deterioration information selected from the group consisting of predicted charge / discharge characteristics, predicted electrochemical analysis data, and predicted material analysis data after the deterioration of the storage battery; simplifying an association model that previously associates the causal relationships between the life-determining event, the plurality of deterioration events, and a plurality of candidate operable factors by excluding one or more deterioration events whose deterioration level is lower than a threshold, and as a result, excluding any deterioration events and candidate operable factors that have lost their causal relationship with the life-determining event; and outputting one or more candidate operable factors that are directly or indirectly causally related to the life-determining event in the simplified association model as operable factors that affect the life of the storage battery. A program for predicting operable factors, which is characterized by causing a computer to execute the above.

Images