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

Abstract

To provide a technology capable of predicting abnormality of a storage battery with high accuracy by using data of the battery measured in periodically inspecting a storage battery system.SOLUTION: A program causes a computer to function as each means of an information processing device for detecting an abnormality sign relating to a plurality of storage batteries constituting an uninterruptible power supply. The information processing device includes acquisition means for acquiring a dataset, for a specific storage battery, to be recorded in each periodical inspection performed at a prescribed time interval and including information on a measurement result to the specific storage battery and information on a use state of the specific storage battery, and determination means for determining the existence / nonexistence of an abnormality sign of the specific storage battery by inputting the dataset for the specific storage battery to a first learning model which has learned in advance. The information on the measurement result to the specific storage battery includes a change degree of an internal resistance value of the specific storage battery in each prescribed time interval, and the information on the use state of the specific storage battery includes the number of use days from a charging date.SELECTED DRAWING: Figure 3

Description

Technical Field

[0001] The present invention relates to an information processing apparatus, an information processing method, a program, and a data structure.

Background Art

[0002] Conventionally, in a system using a power storage module in which a plurality of power storage cells are connected in series, a technique for determining deterioration of the power storage cells is known (Patent Document 1). In the technique proposed in Patent Document 1, a learning model outputs the determination of deterioration of the power storage cell and the presence or absence of environmental abnormality based on power storage element information, measured value time-series data, and predicted value time-series data of the power storage cell.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problem to be Solved by the Invention

[0004] Meanwhile, for example, in a data center, a storage battery system as an uninterruptible power supply may be used to enable continuous operation of computers installed in the data center. The storage battery system as an uninterruptible power supply charges a plurality of storage batteries, and supplies power from the storage batteries to facilities in the data center when a power failure occurs.

[0005] Some storage battery systems have a configuration in which, for example, 100 or more storage batteries are connected in series. In such a storage battery system, if an abnormality occurs in one storage battery and a disconnection occurs, there is a risk of losing the power supply function, so it is necessary to detect signs of abnormality in the storage battery at an early stage and perform appropriate maintenance such as replacement.

[0006] On the other hand, in countries where power outages are infrequent, battery storage systems rarely become operational during actual power outages. Nevertheless, configuring a battery storage system to constantly monitor the status of each battery with sensors would make the system complex and expensive. In some cases, even without constant monitoring by sensors, the battery storage system can be properly maintained by performing regular manual inspections of the batteries at predetermined time intervals (e.g., every month to several months).

[0007] Given this background, there is a need for technology that can accurately predict battery malfunctions by utilizing battery data measured during periodic inspections in battery storage systems.

[0008] This invention has been made in view of the above problems, and its objective is to realize a technology that can accurately predict abnormalities in a battery using battery data measured during periodic inspections of the battery system. [Means for solving the problem]

[0009] To solve this problem, for example, the program of the present invention has the following configuration. That is, A program that causes a computer to function as one of the means of an information processing device for detecting abnormal signs related to multiple batteries constituting an uninterruptible power supply, wherein the information processing device is Acquisition means for acquiring a dataset for a specific battery, which includes information on measurement results for a specific battery and information on the usage status of the specific battery, recorded at each periodic inspection performed at predetermined time intervals, wherein the specific battery is one of a plurality of batteries, including batteries from different manufacturers or of different models. For the dataset for the aforementioned specific battery, The process is normalized according to the manufacturer or model of the aforementioned specific storage battery. Processing means to apply the processing, The system includes a determination means that determines whether or not there are signs of abnormality in the specific battery, using a pre-trained learning model that takes as input a dataset for the specific battery to which the above processing has been applied, The information relating to the measurement results for the specific storage battery includes the internal resistance value of the specific storage battery for each predetermined time interval and the slope of the internal resistance value of the specific storage battery at each predetermined time interval, and the information relating to the usage status of the specific storage battery includes the number of days used from the charging date. [Effects of the Invention]

[0010] According to the present invention, it becomes possible to accurately predict abnormalities in a battery by using battery data measured during periodic inspections of the battery system. [Brief explanation of the drawing]

[0011] [Figure 1] This figure illustrates an overview of the battery abnormality prediction system according to an embodiment of the present invention. [Figure 2A] This diagram schematically shows an example of the configuration of a battery group within a data center according to this embodiment. [Figure 2B] This diagram schematically shows an example of the battery configuration in a specific battery group according to this embodiment. [Figure 2C] This diagram illustrates the relationship between changes in the internal resistance of a battery and battery malfunctions. [Figure 3] Figure showing an example of training data according to this embodiment. [Figure 4] Figure showing another example of training data according to this embodiment. [Figure 5] This figure shows an example of the functional configuration of the information processing device according to this embodiment. [Figure 6] This figure shows an example of the functional configuration of the communication device according to this embodiment. [Figure 7A] This figure shows an example of a dataset stored in the measurement data DB according to this embodiment. [Figure 7B] This figure shows an example of a dataset input to the learning model during the estimation phase according to this embodiment. [Figure 8] Diagram showing an example of the relationship between the timing of periodic inspections and discharge time according to the present embodiment [Figure 9A] Flowchart showing a series of operations in the learning stage by the information processing apparatus according to the present embodiment [Figure 9B] Flowchart showing a series of operations in the estimation stage by the information processing apparatus according to the present embodiment [Figure 10] Diagram showing an example of a notification screen for abnormality signs according to the present embodiment [Figure 11] Schematic diagram showing the data distribution of a plurality of storage batteries according to the present embodiment [Mode for Carrying Out the Invention]

[0012] Hereinafter, embodiments will be described in detail with reference to the accompanying drawings. Note that the following embodiments do not limit the invention according to the claims, and not all combinations of features described in the embodiments are necessarily essential to the invention. Two or more of the plurality of features described in the embodiments may be arbitrarily combined. In addition, the same or similar components are denoted by the same reference numerals, and redundant descriptions are omitted.

[0013] (Overview of storage battery abnormality sign detection system) The overview of the storage battery abnormality sign detection system will be described with reference to FIG. 1. The storage battery abnormality sign detection system 10 includes an information processing apparatus 100 which is an information processing server, and a communication device 210. The communication device 210 is, for example, a personal computer or a mobile terminal disposed in the data center 200. The communication device 210 accepts data input of measurement results related to storage batteries and usage status from a maintenance person in charge of the storage battery group 230, and transmits the input information related to the measurement results and usage status to the information processing apparatus 100 via a network. Further, the communication device 210 may accept data input from the maintenance person in charge of the storage battery group 230 or the administrator of the storage battery abnormality sign detection system 10, and transmit correct data, which will be described later, in the learning data to the information processing apparatus 100.

[0014] The information processing device 100 receives and records information regarding measurement results and usage status from the communication devices 210 of multiple data centers, and detects signs of battery abnormalities based on a machine learning model.

[0015] The data center 200 is home to a group of servers 220 that perform arbitrary processing (for example, processing for tenants that use the server resources of the server group). Under normal circumstances, the server group 220 operates on power supplied from a power source (not shown) within the data center 200. The battery group 230 includes multiple batteries and functions as an uninterruptible power supply, providing power to the server group 220 in emergencies such as power outages.

[0016] The battery cluster 230 within the data center 200 is composed of multiple battery systems (referred to here as UPS battery systems), as shown in Figure 2A, for example. Each of the multiple UPS battery systems is distinguished as, for example, "UPS No. 1," "UPS No. 2," "UPS No. 3," and "UPS Backup," and can independently supply power to the server cluster 220. The "UPS Backup" system is activated, for example, in the event of failure or maintenance of at least one of "UPS No. 1," "UPS No. 2," and "UPS No. 3," or in the event that the required power cannot be supplied due to insufficient power supply.

[0017] Each UPS battery system includes multiple battery groups 240, such as battery group A and battery group B. These battery groups 240 may be connected in series. A battery group 240 is configured to generate the required voltage by connecting multiple batteries 250 in series, for example, as shown in Figure 2B. If a malfunction such as a broken wire occurs in one battery 250, the UPS battery system may lose its function as a UPS power supply. Therefore, accurately detecting signs of malfunction in each battery is an important issue for the UPS power supply to function properly.

[0018] (Detection of signs of battery malfunction) Refer to Figure 2C to explain the detection of abnormal signs in storage batteries. Figure 2C shows an example of the changes in the internal resistance value of several individual storage batteries. In Figure 2C, the horizontal axis represents time, and the vertical axis represents the internal resistance value (mΩ). Graph 261 shows the changes in the internal resistance value measured at times t-3, ..., t, and t+1 for a specific storage battery. Times t-3, ..., t, and t+1 correspond to the timing of measurements during periodic inspections performed in sequence. In other words, Graph 261 is a graph of the measurement results for storage batteries recorded during periodic inspections. Graphs 262 and 263 also show the measurement results for individual storage batteries.

[0019] Conventional anomaly detection methods include determining that a battery's internal resistance exceeds a predetermined threshold of 264, thereby indicating an impending anomaly, and subsequently replacing or inspecting the battery. However, as shown in Graph 262, while the threshold was not exceeded at inspection timing t, if the battery's internal resistance increases rapidly between inspection timing t and t+1, there is a risk that the battery may fail before inspection timing t+1.

[0020] To address these challenges, this embodiment detects batteries that will exceed an internal resistance threshold in the future. In this embodiment, in addition to the internal resistance value, the trend of changes in the internal resistance value is taken into consideration to detect signs of battery abnormality. In fact, in a normal battery, even if the internal resistance value increases over time, the rate of increase decreases. In contrast, the internal resistance value of a battery experiencing an abnormality increases exponentially (both the internal resistance value and the rate of increase increase). This embodiment focuses on this trend of changes in internal resistance value.

[0021] Furthermore, in the abnormality prediction detection according to this embodiment, a machine learning model is trained using data that can take into account the changes in the internal resistance value as training data, and the presence or absence of abnormality is determined by the machine learning model. For example, in the example in Figure 2C, data at times t, t-1, and t-2 are input to the (trained) machine learning model, and it is determined whether an abnormality is indicated (exceeds threshold 264) by time t+1. It is determined that there is no abnormality for the batteries corresponding to graphs 261 and 263, and that there is an abnormality for the battery corresponding to graph 262.

[0022] Figure 3 shows an example of the training data (also called training data) used in this embodiment. This training data is stored in the training data DB of the information processing device 100, which will be described later. The training data includes, for example, a battery identification ID, information on the measurement results corresponding to time t-2 to time t, information on usage status, and a judgment.

[0023] The "Judgment" represents the correct data (also called a label), and based on past periodic inspections and the fact of abnormal occurrences, batteries that exceeded the threshold of 264 between time t and t+1 are marked with "NG" (indicating an abnormality). On the other hand, batteries that did not exceed the threshold of 264 between time t and t+1 are marked with "OK" (indicating no abnormality occurred).

[0024] The battery identification ID is an ID used to identify each battery. Beyond simply identifying a battery, it may also be possible to specify, for example, which data center the battery is located in, which UPS battery system it belongs to, and which specific battery within that UPS system it is. This allows for training the model using training data specific to a particular data center, or training the model using training data related to a battery in a specific UPS battery system. If different types of batteries (i.e., batteries manufactured by different companies or batteries of different characteristics) are deployed in each data center or UPS system, it becomes possible to train the model for each battery with different characteristics, thereby realizing a learning model suitable for each type of battery.

[0025] The measurement data corresponding to time t-2 includes the internal resistance value (mΩ) of the target battery at time t-2 and the slope of the internal resistance value between time t-1 and time t-2. The slope of the internal resistance value may be, for example, the difference in internal resistance values ​​between time t-1 and time t-2, or it may be the difference divided by the number of days between time t-1 and time t-2. Furthermore, these values ​​(such as the difference divided by the number of days between time t-1 and time t-2) may be multiplied by a predetermined scaling factor.

[0026] The measurement data corresponding to time t-1 includes the internal resistance value (mΩ) of the target battery at time t-1 and the slope of the internal resistance value between time t and time t-1. The slope of the internal resistance value may be, for example, the difference in internal resistance values ​​between time t and time t-1, or it may be the value obtained by dividing that difference by the number of days between time t and time t-1. Furthermore, these values ​​may be multiplied by a predetermined scaling factor.

[0027] By adding the slope of the internal resistance to the measurement data at time t-2 and time t-1, the progression of the rate of change in the internal resistance can be explicitly provided to the machine learning model. In other words, unlike when the training data simply includes the internal resistance, in this embodiment, the occurrence of a battery abnormality can be predicted by explicitly taking into account the progression of the rate of change in the internal resistance (i.e., the accelerating increase in the internal resistance).

[0028] The measurement data corresponding to time t includes the internal resistance value (mΩ) of the battery in question at time t. The data indicating usage status includes information on the period of use, and the information on the period of use indicates the number of days elapsed since the first charging of the battery (number of days used since charging). It has been found that the number of days elapsed since the start of battery use correlates with the occurrence of abnormalities.

[0029] By using this type of training data, it becomes possible to predict battery degradation trends by leveraging the results of past periodic inspections, thereby enabling necessary battery countermeasures to be implemented before abnormalities occur. Furthermore, it becomes possible to predict battery abnormalities where the internal resistance value changes rapidly between time t (the current inspection time) and time t+1 (the next inspection time) by explicitly taking into account the progression of the degree of change in the internal resistance value.

[0030] In this embodiment, we will explain using xgboost, a gradient boosting method based on the decision tree algorithm, as an example of a machine learning model. However, the machine learning model may be constructed using other machine learning models, for example, a deep neural network. In addition to using a deep neural network with a simple configuration of hidden layers, a neural network using an RNN (recurrent neural network) such as LSTM, which can more appropriately incorporate the characteristics of temporal changes in time series data, may also be used. Furthermore, a deep neural network having an attention mechanism (also called an attention mechanism) that can provide an output that focuses on characteristic data in the time series data may also be used.

[0031] The training data may include further information, as shown in Figure 4. It may include other measured values ​​such as the battery voltage (V) at time t-2, time t-1, and time t. In addition, information regarding usage may include connection location information in addition to information on the period of use. The connection location information is the connection location (order from the front) of the target battery in a group of batteries connected in series. As mentioned above, multiple batteries are arranged in series within a battery group. The batteries are arranged in a two-dimensional or three-dimensional manner, for example, in a rack or box. Therefore, when the batteries discharge during a power outage or periodic inspection, the temperature changes according to the arrangement in the box affect the degradation of the batteries. However, the timing of battery measurements during periodic inspections may not be when the batteries are discharging, as shown in Figure 8. For example, battery discharge is expected to occur for a very limited time during inspection work for operational verification, or when power supply is needed due to a power outage. Therefore, the timing of measurement and recording during periodic inspections may not allow for the measurement of temperatures that would affect the batteries. Therefore, in this embodiment, the battery connection position (or the arrangement of batteries in a two-dimensional or three-dimensional configuration) is used as an input parameter for the learning model. Alternatively, discharge time may be used instead of, or in addition to, the battery connection position. Discharge time is the cumulative time over which the battery has discharged. This makes it possible to reflect the effects of battery operation in the detection of degradation precursors.

[0032] In this embodiment, the case where the communication device 210 transmits information regarding measurement results, usage status, and correct answer data to the information processing device 100, and the information processing device 100 generates learning data, will be described as an example. However, the communication device 210 may generate the learning data according to this embodiment and transmit the generated learning data to the information processing device 100. In this case, the information processing device 100 only needs to store the acquired learning data in the learning data DB 522.

[0033] (Configuration of information processing device) Next, an example of the functional configuration of the information processing device 100 will be described with reference to Figure 5. Note that each of the functional blocks described with reference to the following figures may be integrated or separated, and the functions described may be implemented in other blocks. Furthermore, what is described as hardware may be implemented in software, and vice versa.

[0034] The communication unit 501 includes a communication circuit or communication module that communicates via a network with a communication device 210 used by battery maintenance personnel. The communication unit 501 communicates with multiple communication devices 210.

[0035] The control unit 502 includes a central processing unit, the CPU 510, and RAM 511. The CPU 510 may consist of one or more CPUs. The control unit 502 executes computer programs stored in the recording unit 504 to perform battery abnormality prediction detection processing, which will be described later, and to control the operation of each part of the information processing device 100. The control unit 502 may further include a computing unit (e.g., a GPU) or dedicated hardware for performing statistical processing such as machine learning at a higher speed.

[0036] RAM 511 is a volatile storage medium such as DRAM, and it temporarily stores parameters and processing results for the control unit 502 to execute a computer program.

[0037] The power supply unit 503 is a circuit or module that provides power for the operation of each part of the information processing device 100. The power supply unit 503 may also be configured to include a battery.

[0038] The recording unit 504 includes, for example, a non-volatile recording medium such as a hard disk or semiconductor memory, and records setting values ​​and calculation results necessary for the operation of the information processing device 100. The recording unit 504 also stores information regarding measurement results and usage status received from the communication device 210 in the measurement data DB 521. Furthermore, the recording unit 504 stores the aforementioned learning data in the learning data DB 522.

[0039] The battery data receiving unit 512 receives information regarding measurement results and usage status entered by the battery maintenance personnel from the communication device 210 and records it in the measurement data DB 521 of the recording unit 504. The information regarding measurement results and usage status stored in the measurement data DB 521 is configured as shown in Figure 7A, for example.

[0040] The data shown in Figure 7A includes a battery identification ID 701, information on measurement results and usage status for each hour, and information on usage status that is constant with respect to time 704. The information on measurement results and usage status for each hour includes, for example, information on measurement results and usage status measured at time tn 702, information on measurement results and usage status measured at time tn-1, and information on measurement results and usage status measured at time t 703.

[0041] The battery identification ID 701 is the same as described above in Figures 3 and 4. The internal resistance and voltage values ​​in the information regarding the measurement results for each hour are the same as the internal resistance and voltage values ​​described above in Figure 4. The date is the date on which the data such as the internal resistance was measured. The usage period and discharge time in the information regarding usage status represent the cumulative number of days up to the corresponding time (e.g., tn). The information regarding usage status 704 includes information on the connection location, and this connection location information is the same as the connection location information shown in Figure 4.

[0042] Refer to Figure 5 again. The learning data generation unit 513 generates the above-mentioned learning data using data from a specific battery (specified, for example, by the battery maintenance person or the administrator of the abnormality prediction detection system 10) from the data stored in the measurement data DB 521. For example, the learning data generation unit 513 calculates the slope of the internal resistance value at time t-1 from the date and internal resistance value at time t-1 in the measurement data DB 521 and the date and internal resistance value at time t-1. It also generates information on usage status as shown in Figure 3 or Figure 4 based on information on usage status at time t and information on usage status that does not change over time 704. A specific battery is a battery that has actually experienced an abnormality or has been confirmed to have been operating normally. According to the inventor's experiments, even if the learning data consists of, for example, data from three periodic inspections (times t-2, t-1, and t), abnormality prediction can be detected well. However, the system is not limited to this example, and data from more periodic inspections, such as five times, may be used.

[0043] The learning data generation unit 513 adds learning data judgment information (i.e., correct answer data indicating "OK" or "NG") to the battery identification ID data of the specific battery. The learning data generation unit 513 receives correct answer data via the communication device 210, for example, from a battery maintenance person or the administrator of the battery abnormality prediction detection system 10, and adds it to the learning data. The learning data generation unit 513 stores the generated learning data in the learning data DB 522.

[0044] The measurement data acquisition unit 514 generates information about measurement data and usage status for a specific battery from the measurement data DB 521 in order to perform the anomaly prediction detection process in the estimation stage described later. The batteries that the measurement data acquisition unit 514 processes are those that are subject to determination of the presence or absence of anomalies. The measurement data acquisition unit 514 generates a dataset to be input into the learning model for anomaly prediction detection by performing the same processing as the data generation process other than "determination" described for the learning data generation unit 513. As shown in Figure 7B, the dataset generated by the measurement data acquisition unit 514 includes the internal resistance value and the slope of the internal resistance value at time t-2 and time t-1, respectively, the internal resistance value at time t, and information on the usage period. Note that the example dataset shown in Figure 7B is in the same format as the learning data shown in Figure 3. Furthermore, when generating a dataset in a format corresponding to the training data shown in Figure 4, the generated dataset will include the internal resistance value, the slope of the internal resistance value, and the voltage at time t-2 and time t-1, respectively, as well as the internal resistance value and voltage at time t, and information on the usage period, connection location, and discharge time.

[0045] The preprocessing unit 515 performs common preprocessing on the training data and the data shown in Figure 7B as a preliminary step before inputting the data into the learning model. This preprocessing reduces the impact of variations in characteristics on the learning model's learning and estimation results when batteries are manufactured by multiple manufacturers and of different models, and the characteristics of the batteries vary depending on the manufacturer and model. The preprocessing may include, for example, a process to normalize each parameter of the information regarding measurement results and information regarding usage conditions. Normalization may involve scaling each parameter so that it falls within the range of 0 to 1 (for example, by using the maximum value for each manufacturer and model). Alternatively, the preprocessing unit 515 may use, for example, a value obtained by dividing the measurement result, such as voltage value or internal resistance value, by the equipment specification information. For example, by dividing the measured voltage value by the standard voltage, which differs for each battery product, the ratio to the standard voltage is calculated, and the calculated ratio to the standard voltage is input into the learning model. The standard voltage differs depending on the number of series cells in the product. Furthermore, the pre-processing unit 515 may calculate a ratio to the standard internal resistance by dividing the internal resistance value by a standard internal resistance value which differs for each product, and use this ratio. The standard internal resistance value varies depending on the number of series cells and terminal structure within the product. Alternatively, the battery's service life may be divided by the average lifespan which differs for each product, and a ratio to the average lifespan may be calculated and used. Note that information such as the maximum value used for normalization, the standard voltage for each manufacturer and model, the standard internal resistance value, and the average lifespan may be pre-recorded in the recording unit 504.

[0046] The model processing unit 516 takes the training data shown in Figure 3 or Figure 4 as input and trains its internal machine learning model. It also uses the trained machine learning model to take the input data shown in Figure 7B and executes anomaly prediction detection processing for the storage battery. The machine learning model used by the model processing unit 516 in this embodiment functions as a classifier that outputs the prediction result as binary (OK or NG). As described above, in this embodiment, the training model is, for example, xgboost, a gradient boosting method based on the decision tree algorithm. However, the machine learning model may be configured using, for example, a deep neural network. The anomaly information provision unit 517 transmits the results of the anomaly prediction detection processing in the estimation stage by the model processing unit 516 to the communication device 210. The anomaly information provision unit 517 transmits the results of the anomaly prediction detection processing to the communication device 210 as information for the anomaly prediction notification screen, which will be described later with reference to Figure 10.

[0047] (Configuration of communication equipment) Next, an example of the functional configuration of the communication device 210 will be described with reference to Figure 6. In this embodiment, a personal computer is used as an example of a communication device, but the communication device may be other electronic devices such as a smartphone or tablet terminal. Note that each of the functional blocks described with reference to the following figures may be integrated or separated, and the functions described may be implemented in other blocks. Also, what is described as hardware may be implemented in software, and vice versa.

[0048] The communication unit 601 includes communication circuits and other components, and communicates with the information processing device 100 by, for example, connecting to the Internet via wireless communication such as a wireless LAN, or connecting to a network via wireless communication for mobile phones.

[0049] The control unit 602 includes a CPU 610 and RAM 611, and controls the operation of each part of the communication device 210 by having the CPU 610 execute a computer program recorded in the recording unit 607, for example.

[0050] The operation unit 603 includes at least one of a keyboard, mouse, or touch panel, and can perform operations on the GUI for various operations displayed on the display unit 606 (for example, inputting information about measurement results and usage status, and providing correct answer data). The power supply unit 604 provides power to each part of the communication device 210. The imaging device 305 is, for example, a camera mechanism including an image sensor, and takes images in response to instructions from the control unit 602.

[0051] The display unit 606 includes, for example, a display device such as an LCD or OLED. The display unit 606 displays a GUI for inputting information about measurement results and usage status, and for assigning correct answer data, as well as a GUI for an abnormality prediction notification screen, etc., in response to instructions from the control unit 602.

[0052] The recording unit 607 includes, for example, non-volatile memory such as semiconductor memory, and stores programs and settings executed by the control unit 602. The computer programs stored in the recording unit 607 include an operating system and various applications for realizing the functions of the communication device 210.

[0053] (A series of operations for anomaly prediction detection during the learning phase) Next, a series of operations for anomaly prediction detection processing during the learning phase, which is performed in the information processing device 100, will be explained with reference to Figure 9A. This processing is realized by the CPU 510 of the control unit 502 executing a computer program recorded in the recording unit 504. In the following explanation, for the sake of simplicity, the processing entity for each step will be collectively described as the control unit 502, but each part that functions within the control unit 502 will execute the corresponding processing according to the processing content. Note that this processing is the process of training the machine learning model in the model processing unit 516, and the machine learning model has not yet completed training.

[0054] In S401, the control unit 502 receives information regarding the measurement results and information regarding the usage status from the communication device 210. As described above, the control unit 502 stores the information regarding the measurement results and information regarding the usage status in the measurement data DB 521 in the format shown in Figure 7A.

[0055] In S402, the control unit 502 obtains information necessary for generating training data from the communication device 210. The information necessary for generating training data is the correct answer data shown in Figure 3 or Figure 4.

[0056] In S404, the control unit 502 generates the training data shown in Figure 3 or Figure 4 based on the data shown in Figure 7A (information regarding measurement results and information regarding usage status) stored in the measurement data DB 521 and the correct data acquired in S402. The generation of the training data is performed as described above for the training data generation unit 513.

[0057] In S405, the control unit 502 performs training on the learning model. The control unit 502 trains the machine learning model using the generated training data. For training the machine learning model, known training methods can be used for decision trees and deep neural networks, so a detailed explanation is omitted. Once the control unit 502 has finished training the machine learning model using the training data, it terminates the series of operations for anomaly prediction detection processing during the training phase.

[0058] (A series of operations for anomaly prediction detection during the estimation phase) Next, a series of operations for the anomaly prediction detection process in the estimation stage, which is performed in the information processing device 100, will be explained with reference to Figure 9B. This process is realized by the CPU 510 of the control unit 502 executing a computer program recorded in the recording unit 504. In the following explanation, for the sake of simplicity, the processing entity for each step will be collectively described as the control unit 502, but each part that functions within the control unit 502 will execute the corresponding process according to the processing content. This process uses a trained machine learning model in the model processing unit 516 to determine whether or not there is an anomaly prediction from newly input battery data. In the process shown in Figure 9B, the case of processing for one specific battery is explained as an example, but the processes of S921 and S922 may be repeated to perform anomaly prediction detection for multiple batteries.

[0059] In S921, the control unit 502 generates the data shown in Figure 7B from the data shown in Figure 7A stored in the measurement data DB 521 for a specific battery that is the target of the abnormality prediction detection process. After generating the data shown in Figure 7B, the control unit 502 may apply the preprocessing performed by the preprocessing unit 515 described above to the generated data.

[0060] In S922, the control unit 502 executes anomaly prediction detection processing. The control unit 502 uses a trained machine learning model to determine whether or not there are any anomaly signs in the specific battery being processed. The control unit 502 inputs the data generated in S921 into the trained machine learning model and obtains an output from the machine learning model (i.e., a binary value of OK or NG).

[0061] In S923, the control unit 502 transmits information about the battery in which abnormalities have been detected (for example, the battery identification ID and the change in the degree of change of the internal resistance value) to the communication device based on the judgment result obtained from the machine learning model. After transmitting the battery information to the communication device 210, the control unit 502 terminates this series of processes.

[0062] Alternatively, the processes S921 to S922 may be repeatedly executed for multiple batteries, and information on abnormal signs of multiple batteries (also called an abnormal sign notification screen) as shown in Figure 10 may be transmitted to the communication device 210. The abnormal sign notification screen 1000 shown in Figure 10 displays the battery identification ID of one or more batteries in which abnormal signs have been detected, and the battery information 1002 corresponding to the battery identification ID. The battery information 1002 may also display at least one of the characteristics of the battery showing abnormal signs, such as the change in the slope of the internal resistance value 1003, the usage period 1004, and the connection position 1005. In this way, batteries that are estimated to have abnormal signs at the current time (time t) can be grasped at once, and the characteristics of each battery can be viewed at a glance, making it easy to confirm the validity of the estimation results.

[0063] As described above, in this embodiment, a dataset for a specific battery, including information on measurement results and usage status recorded for each periodic inspection performed at predetermined time intervals, is acquired and input into a trained machine learning model to determine whether or not there are signs of abnormality in the specific battery. At this time, the information on measurement results for the specific battery includes the degree of change in the internal resistance value of the specific battery at predetermined time intervals, and the machine learning model is explicitly provided with the trend of the degree of change in the internally provided value. In this way, it becomes possible to accurately predict battery abnormalities using battery data measured during periodic inspections of the battery system. According to this embodiment, even with a small amount of periodic inspection data, such as three times' worth, battery abnormalities can be accurately predicted by providing the trend of the degree of change in the internally provided value, so the configuration of the battery system can be simplified (without requiring sensors that constantly monitor the characteristics of the battery).

[0064] In the above embodiment, internal resistance, slope of internal resistance, and voltage were given as examples of information regarding the measurement results of the storage battery, but the information regarding the measurement results may include other information as well. For example, it may include information such as the temperature of the storage battery, the number of discharge cycles, the discharge time, or at least one of the discharge capacity obtained during periodic inspections.

[0065] Furthermore, in the above-described embodiment, the abnormality prediction detection process determines whether or not there are abnormalities for each individual battery. However, in addition to the abnormality prediction determination for each individual battery as described above, the model processing unit 516 may use machine learning algorithms such as clustering or SVM (Support Vector Machine) to determine batteries that fall outside the class of batteries determined to be normal among the multiple batteries as batteries showing abnormalities.

[0066] For example, as shown in Figure 11, if we consider a space where the components of a vector are information about the measurement results and usage status of each battery, the data of multiple batteries will be distributed at various positions in the space (e.g., 1103 and 1104). By applying machine learning algorithms such as clustering and SVM to this battery data, we can distinguish between battery 1103, which belongs to the normal battery class 1102, and other batteries 1104. Then, the distinguished battery 1104 is determined to be a battery showing signs of abnormality.

[0067] The control unit 502 may execute multiple machine learning algorithms using the model processing unit 516, and if the battery in which abnormality signs were detected by the processing in S922 matches the battery 1104 which does not belong to the class of normal batteries, it may ultimately detect the battery as having abnormality signs. In this way, maintenance can be performed on batteries with a high probability of having abnormality signs, as determined by multiple machine learning algorithms. On the other hand, the control unit 502 may execute multiple machine learning algorithms using the model processing unit 516, and if any one of the machine learning algorithms determines that there are abnormality signs, it may ultimately detect the battery as having abnormality signs. In this case, maintenance can be performed on a wide range of batteries that may have abnormality signs.

[0068] The invention is not limited to the embodiments described above, and various modifications and changes are possible within the scope of the gist of the invention. [Explanation of symbols]

[0069] 100... Information processing device, 210... Communication device, 513... Learning data generation unit, 514... Measurement data acquisition unit, 515... Preprocessing unit, 516... Model processing unit

Claims

1. A program that causes a computer to function as one of the means of an information processing device for detecting abnormal signs related to multiple batteries constituting an uninterruptible power supply, wherein the information processing device is Acquisition means for acquiring a dataset for a specific battery, which includes information on measurement results for a specific battery and information on the usage status of the specific battery, recorded at each periodic inspection performed at predetermined time intervals, wherein the specific battery is one of a plurality of batteries, including batteries from different manufacturers or of different models. Processing means for applying a normalization process to a dataset for the aforementioned specific battery according to the manufacturer or model of the aforementioned specific battery, The system includes a determination means that determines whether or not there are signs of abnormality in the specific battery, using a pre-trained learning model that takes as input a dataset for the specific battery to which the above processing has been applied, A program characterized in that the information relating to the measurement results for the specific storage battery includes the internal resistance value of the specific storage battery for each predetermined time interval and the slope of the internal resistance value of the specific storage battery at each predetermined time interval, and the information relating to the usage status of the specific storage battery includes the number of days used since the charging date.

2. The program according to claim 1, characterized in that the process normalizes the internal resistance value included in the information regarding the measurement result using a value predetermined for each manufacturer.

3. The program according to claim 1, characterized in that the processing normalizes the internal resistance value included in the information regarding the measurement result using a value predetermined for each model.

4. The program according to claim 2 or 3, characterized in that the process normalizes the internal resistance value included in the information regarding the measurement result using the maximum value of the internal resistance value predetermined for each manufacturer or model.

5. The program according to claim 1, wherein the information relating to the usage status of the particular battery further includes the connection position of the particular battery among the plurality of batteries connected in series.

6. The program according to claim 1, wherein the information relating to the usage status of the particular storage battery further includes the arrangement of the particular storage battery among the plurality of storage batteries arranged in two dimensions or three dimensions.

7. The program according to claim 1, wherein the information relating to the measurement results for the specific storage battery further includes at least one of the number of discharges, the discharge time, or the discharge capacity.

8. An information processing method in which each step is executed in an information processing device for detecting abnormal signs related to multiple storage batteries constituting an uninterruptible power supply, The method involves obtaining a dataset for a specific battery, which includes information on measurement results for that specific battery and information on the usage status of that specific battery, recorded at each periodic inspection performed at predetermined time intervals, wherein the specific battery is one of the multiple batteries, including batteries from different manufacturers or of different models. Applying a normalization process to the dataset for the aforementioned specific battery according to the manufacturer or model of the aforementioned specific battery, This includes determining whether or not there are any signs of abnormality in the specific battery using a pre-trained learning model that takes as input the dataset for the specific battery to which the above processing has been applied, The information processing method is characterized in that the information relating to the measurement results for the specific storage battery includes the internal resistance value of the specific storage battery for each predetermined time interval and the slope of the internal resistance value of the specific storage battery at each predetermined time interval, and the information relating to the usage status of the specific storage battery includes the number of days used since the charging date.

9. An information processing device for detecting abnormal signs related to multiple storage batteries constituting an uninterruptible power supply, Acquisition means for acquiring a dataset for a specific battery, which includes information on measurement results for a specific battery and information on the usage status of the specific battery, recorded at each periodic inspection performed at predetermined time intervals, wherein the specific battery is one of a plurality of batteries, including batteries from different manufacturers or of different models. Processing means for applying a normalization process to a dataset for the aforementioned specific battery according to the manufacturer or model of the aforementioned specific battery, The system includes a determination means that determines whether or not there are signs of abnormality in the specific battery, using a pre-trained learning model that takes as input a dataset for the specific battery to which the above processing has been applied, The information processing device is characterized in that the information relating to the measurement results for the specific storage battery includes the internal resistance value of the specific storage battery for each predetermined time interval and the slope of the internal resistance value of the specific storage battery in each predetermined time interval, and the information relating to the usage status of the specific storage battery includes the number of days used since the charging date.

Images