Abnormality detection system
The anomaly detection system addresses inefficiencies in data center cooling by updating parameters based on historical data and similarity with other server rooms, enhancing energy efficiency.
Patent Information
- Application Number
- JP2023077333
- Authority / Receiving Office
- JP · JP
- Patent Type
- Applications
- Current Assignee / Owner
- Filing Date
- 2023-05-09
- Publication Date
- 2024-11-21
AI Technical Summary
Existing anomaly detection systems in data centers struggle to accurately adjust parameters for cooling devices due to varying factors such as IT equipment load, outside temperature fluctuations, and building characteristics, leading to inefficiencies in power consumption that are not optimally detected.
An anomaly detection system that includes processors and storage devices to collect and analyze data history from multiple sources, update anomaly detection parameters based on similarity with other server rooms, and adjust settings to optimize cooling device performance.
The system effectively detects anomalies in cooling device power consumption, allowing for more accurate parameter adjustments and improved energy efficiency in data centers.
Smart Images

Figure 2024162111000001_ABST
Abstract
Description
[Technical field]
[0001] The present invention relates to an anomaly detection system. [Background technology]
[0002] In recent years, energy conservation has been required in order to achieve the Sustainable Development Goals (SDGs). At the same time, the demand for data centers, which consume a large amount of the world's energy in modern times, is continually increasing, and there is a need to improve the power efficiency of data centers. For data center operators, the main source of electricity that can be reduced is air conditioners (air conditioners), which include cooling devices such as chillers.
[0003] While this reduction in power consumption is important, the power consumption of cooling equipment is affected by the load of IT equipment and fluctuations in outside temperature, as well as the characteristics of the building and server room itself, and the way in which it is affected varies from server room to server room in a data center. Furthermore, these characteristics change over time depending on the IT equipment contained in the server room, the workload, seasonal fluctuations, etc. Therefore, although it is necessary to accurately detect inefficiencies in the power consumption of cooling equipment, it is necessary to adjust judgment parameters for each server room in order to properly judge whether or not the power fluctuations are inefficient.
[0004] In order to capture the time-series changes in the characteristics of each server room and adjust the parameters to suit the characteristics of each server room, it is necessary to detect the appropriate feature change points and to appropriately update the anomaly detection conditions. Hereinafter, the anomaly detection conditions are called parameters. Patent Document 1 discloses a method of creating a power consumption prediction model from past data and determining whether or not to update the prediction model based on the difference between the predicted and actual values. [Prior art documents] [Patent documents]
[0005] [Patent Document 1] U.S. Patent No. 10,896,478 Summary of the Invention [Problem to be solved by the invention]
[0006] For example, consider inefficiency detection regarding the amount of power consumed by cooling devices (cooling device power consumption) in each server room in a data center. Inefficiency here refers to, for example, excessive cooling causing the actual room temperature to be lower than the desired room temperature overall or locally within a server room. Inefficiency detection is one type of anomaly detection. Reasons for changes in the characteristics of a server room include an increase / decrease or change in the number of IT devices in the server room, an increase / decrease or change in cooling devices, and manual changes to the settings of cooling devices.
[0007] On the other hand, manual changes to cooling device settings are not necessarily optimal and may contain errors. In addition, depending on the mechanical characteristics of the cooling device, the characteristics of fluctuations in the power consumption of the cooling device may change significantly due to changes in the outside temperature, humidity, room temperature, and power consumption of IT devices (power consumption of IT devices, etc.). Such events can also occur in data from other types of data sources. Therefore, technology that can appropriately detect anomalies in response to changes in the characteristics of the target data is desired. [Means for solving the problem]
[0008] An anomaly detection system according to one embodiment of the present invention includes one or more processors and one or more storage devices, wherein the one or more storage devices store data history for each of a plurality of data sources and anomaly detection parameter history indicating a history of anomaly detection parameters for detecting anomalies in the data for each of the plurality of data sources, and the one or more processors acquire current data of a first data source in the plurality of data sources, and when a relationship between the current data and the current anomaly detection parameters of the first data source satisfies a predetermined update condition, extract one or more periods from the data history of at least a portion of the plurality of data sources that are determined to be similar to a period including the current data by a predetermined method, acquire anomaly detection parameters for the one or more periods from the anomaly detection parameter history of at least a portion of the plurality of data sources, and update the current anomaly detection parameters based on the acquired anomaly detection parameters for the one or more periods. Effect of the Invention
[0009] According to one aspect of the present invention, anomalies in target data can be detected more appropriately. [Brief description of the drawings]
[0010] [Figure 1] Overall configuration of the computing system [Diagram 2] Management computer configuration diagram [Diagram 3] Data store configuration diagram [Figure 4] FIG. 13 is a diagram showing an example of a server room management table; [Diagram 5] FIG. 13 is a diagram showing an example of a table for storing a log of power consumption of a cooling device. [Figure 6] A diagram showing an example of a table for storing logs of power consumption of IT devices, etc. [Figure 7] FIG. 13 is a diagram showing an example of a table for storing logs of sensors, etc. [Figure 8] FIG. 1 is a diagram showing an example of a table for storing a log of meteorological data provided by the Japan Meteorological Agency, etc. [Figure 9]FIG. 13 is a diagram showing an example of a table for storing parameters to be updated. [Figure 10] Flowchart showing data collection process [Figure 11] Flowchart showing a method for determining a variation type in a cooling device power consumption log [Figure 12] Flow showing how to update parameters [Figure 13] Flow showing how to calculate similarity with other server rooms [Figure 14] Example screen output showing the results of a parameter update calculation [Figure 15] Example screen output showing the results of a parameter update calculation DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS EXAMPLES
[0011] A first embodiment of the parameter updating method and the computing device will be described below with reference to Figs. 1 to 15. Fig. 1 is an overall configuration diagram of a computing system 100, which is an example of an anomaly detection system. The computing system 100 includes a first data center 1000 and a computing environment 2000. The first data center 1000 includes a first server room 1000A and a second server room 1000B. The first server room 1000A and the second server room 1000B each include one or more computers 3000, one or more cooling devices 4000, and one or more temperature and humidity sensors 5000.
[0012] The first server room 1000A, the second server room 1000B, and the computing environment 2000 are connected by a wide area network 9000. The server rooms may be geographically distributed. The computing environment 2000 may be an on-premise environment or a public cloud environment. It may also be on a virtual machine.
[0013] In the server rooms 1000A and 1000B, the computer 3000, the cooling device 4000, and the temperature and humidity sensor 5000 are connected by a communication line 8000. The computer 3000, the cooling device 4000, and the temperature and humidity sensor 5000 included in the first server room 1000A and the second server room 1000B are collectively referred to, and no distinction is made in this embodiment.
[0014] The number of data centers included in the computing system 100 is not limited to one, and may be two or more. The first data center 1000 corresponds to a base, and when the number of data centers is two or more, the bases may be geographically dispersed. The number of server rooms included in the first data center 1000 is not limited to two, and may be one, or three or more. The computer 3000 is a device that performs some kind of calculation, and the content of the calculation is not limited. The computer 3000 includes a computer, a data storage device, a network switch, a hub, etc.
[0015] The cooling device 4000 is an air conditioning device that cools the heat generated by the computer 3000 and the heat taken in from the outside air. There are no specific limitations on the configuration, but it may be a device that uses a refrigerant, a chiller that uses cooling water, or a fan. One or more air conditioning devices in the server room make up the air conditioning system of the server room.
[0016] Since the power consumption and heat generation of the computer 3000 increases with an increase in the amount of calculations per unit time, the power consumption of the cooling device 4000 also tends to increase in order to keep the room temperature constant. That is, it can be said that there is a positive correlation between the power consumption of the computer 3000 and the power consumption of the cooling device 4000. Also, since the temperature in the server room rises with an increase in the outside air temperature, the power consumption of the cooling device 4000 also tends to increase in order to keep the room temperature constant. That is, it can be said that there is a positive correlation between the outside air temperature and the power consumption of the cooling device 4000.
[0017] The temperature and humidity sensor 5000 is a device installed in the first server room 1000A or the second server room 1000B. The temperature and humidity sensor 5000 installed in the first server room 1000A observes the temperature and humidity in the first server room 1000A or the second server room 1000B, and the observed data is stored in the data store 7000 via the communication line 8000A or 8000B and the wide area network 9000.
[0018] The temperature and humidity sensor 5000 may be installed anywhere, such as on a wall, on a floor, under a floor, or on a ceiling, but is not limited thereto. The temperature and humidity sensor 5000 may be in the form of a sensor capable of measuring both temperature and humidity, or a sensor capable of measuring only one of them. The temperature and humidity sensor 5000 may be a sensor that measures the temperature or humidity at one point, or a sensor that measures the temperature or humidity locally or entirely in a space.
[0019] The management computer 6000 manages the computer 3000, the cooling device 4000, and the temperature and humidity sensor 5000. The management computer 6000 collects data related to the power consumption of the computer 3000 and the cooling device 4000, and temperature and humidity data measured by the temperature and humidity sensor 5000, and stores the data in a data store 7000. The computer 3000, the management computer 6000, and the data store 7000 may be bare metal, i.e., physical devices, or may be virtual machines. Furthermore, the computer 3000, the management computer 6000, and the data store 7000 may or may not be geographically distributed.
[0020] Although not shown, the computer 3000 , cooling device 4000 , temperature and humidity sensor 5000 , management computer 6000 and data store 7000 may be connected to each other without being connected to the wide area network 9000 .
[0021] 2 is a configuration diagram of a management computer 6000. The management computer 6000 includes a processor 6100, a memory 6200 which is a main storage device, a storage device 6300 which is an auxiliary storage device, a communication interface 6400, and an I / O device 6500. For example, the processor 6100 is a central processing unit, the storage device 6300 includes a non-volatile storage medium, and the memory 6200 includes a volatile storage medium.
[0022] The communication interface 6400 is, for example, a network interface card. The I / O device 6500 is, for example, a mouse, a keyboard, and a liquid crystal display. The I / O device 6500 provides information to the person who uses the management computer 6000 (hereinafter referred to as the "user"), and receives input from the user.
[0023] The storage device 6300 stores working data 6310, power consumption inference result data 6320, and a cooling device power consumption anomaly detection model 6330. The working data 6310 is a part of the data stored in the data store 7000, and is used by the cooling device power consumption fluctuation type determination program 6220, the power consumption prediction program 6230, and the parameter update program 6240. The power consumption inference result data 6320 is a result of inference obtained by executing the power consumption prediction program 6230 using the cooling device power consumption anomaly detection model 6330. In this embodiment, the cooling device power consumption anomaly detection model 6330 is a prediction model for predicting the power consumption of the cooling device, and may be, for example, a machine learning model.
[0024] The output data of the cooling device power consumption anomaly detection model 6330 is the future total power consumption of the cooling devices in the server room. The input data for prediction of the cooling device power consumption anomaly detection model 6330 is, for example, the total power consumption history of the cooling devices 4000 in the server room, the total power consumption history of the computers 3000, the indoor temperature history, and the outdoor temperature history for a specific period. Some of these may be omitted. The cooling device power consumption anomaly detection model 6330 can be trained by an existing method using learning data that is a combination of the above input data and output data.
[0025] The memory 6200 stores a data collection program 6210 , a cooling device power consumption fluctuation type determination program 6220 , a power consumption prediction program 6230 , and a parameter update program 6240 .
[0026] The data collection program 6210 acquires data from the computer 3000 , the cooling device 4000 , and the temperature and humidity sensor 5000 , and stores the data in a cooling device power consumption log table 7320 , an IT equipment power consumption log table 7330 , and a sensor log table 7340 in the data store 7000 .
[0027] Furthermore, weather data for areas geographically close to the first data center 1000 is acquired from sensors installed in the first data center 1000 or from weather information providers, and stored in the weather information log table 7350. In this embodiment, the data includes at least the power consumption data of the cooling device, the power consumption data of the computer, the temperature and humidity measured by sensors installed in the server room, and the outside air temperature and humidity, but is not limited to this and may include other data. In this embodiment, the data collection program 6210 is executed periodically. The processing flow of this program is illustrated in the data storage flow 10010.
[0028] The cooling device power consumption fluctuation type determination program 6220 uses power consumption 7324 acquired from a cooling device power consumption log table 7320 shown in Figures 3 and 5. It determines whether the log shows a continuous change in the power consumption of a cooling device in a specific server room or a rectangular change, and stores the result in the server room table 7310 shown in Figures 3 and 4. In this embodiment, the cooling device power consumption fluctuation type determination program 6220 is executed periodically. The execution flow of the cooling device power consumption fluctuation type determination program 6220 is included in the cooling device power consumption fluctuation type determination flow 11000.
[0029] The power consumption prediction program 6230 predicts the amount of power consumption of the cooling device, and outputs the inference result data 6320 shown in Fig. 2. The processing flow of the inference program itself is not shown, but it is incorporated into the parameter update flow 12000 that is implemented using this.
[0030] In this embodiment, the parameter update program 6240 predicts the amount of power consumption of the cooling device for each server room, judges the necessity of updating the parameters based on the magnitude of the difference between the predicted value and the actual measured value (predicted-actual difference), and updates the parameters if it is judged that a parameter update is necessary. In the example described below, the necessity of parameter update is determined by referring to the parameter. Specifically, the parameter is a threshold value of the predicted-actual difference, and the parameter is updated when the predicted-actual difference exceeds the parameter. The necessity of a parameter update corresponds to the detection of an abnormality that requires a change in the system settings, and the parameter is a parameter for detecting the abnormality. In this embodiment, the parameter update program 6240 is executed periodically.
[0031] In this embodiment, the data stored in the data store 7000 is used to calculate the similarity between server rooms, and similar server rooms are extracted. Specifically, the data used includes power consumption data of cooling devices from a cooling device power consumption log table 7320, power consumption data of IT devices such as computers from an IT device power consumption log table 7330, temperature and humidity measured by a sensor 5000 installed in the server room from a sensor log table 7340, and outdoor temperature and humidity from a weather information log table 7350. The IT device power consumption log table 7330 is shown in Figs. 3 and 6, the sensor log table 7340 is shown in Figs. 3 and 7, and the weather information log table 7350 is shown in Figs. 3 and 8.
[0032] The power consumption prediction program 6230 extracts server room IDs 7311 that are determined to have the same cooling device power consumption fluctuation type 7312 from the server room table 7310. New parameters for the server room are determined using the data stored in the data store 7000 and the parameters 7363 included in the server room table 7310. The processing flow of this program is shown in parameter update flow 12000. Note that in this embodiment, this flow is not executed if the server room is registered in the server room table 7310 and the cooling device power consumption fluctuation type 7312 is Null, but this is not limited to the present embodiment.
[0033] 3 is a configuration diagram of a data store 7000. The data store 7000 includes a processor 7100, a memory 7200 which is a main storage device and stores a data input / output program 7210, a storage device 7300 which is an auxiliary storage device, a communication interface 7400, and an I / O device 7500. For example, the processor 7100 is a central processing unit, the storage device 7300 includes a non-volatile storage medium, and the memory 7200 includes a volatile storage medium.
[0034] The communication interface 7400 is, for example, a network interface card. The I / O device 7500 is, for example, a mouse, a keyboard, and a liquid crystal display. The I / O device 7500 provides information to a user who uses the data store 7000 and receives input from the user.
[0035] The server room table 7310, the cooling device power consumption log table 7320, the IT equipment etc. power consumption log table 7330, the sensor log table 7340, the weather information log table 7350, and the parameter management table 7360 are stored in the storage device 7300. These may be loaded onto and held in the memory 7200.
[0036] The data input / output program 7210 is a program for inputting and outputting data to the server room table 7310, the cooling device power consumption log table 7320, the IT equipment power consumption log table 7330, the sensor log table 7340, the weather information log table 7350, and the parameter management table 7360 via the network. A method (protocol) for inputting and outputting data may be a well-known one, such as a method using an API such as REST or GraphQL, or a method using a data transmission / reception protocol such as NFS, iSCSI, or FTP. The flow of the data input / output program is illustrated in the data storage flow.
[0037] The processor 6100 of the management computer 6000 and the processor 7100 of the data store 7000 each execute a program and operate as a functional unit corresponding to that program. Therefore, a process that has a program as its subject is a process performed by the processor that executes that program or a system that includes that processor.
[0038] 4 is a diagram showing an example of a server room table 7310. The server room table 7310 is made up of a plurality of records, and each record has columns of a server room ID 7311, a cooling device power consumption fluctuation type 7312, a start date 7313, and an end date 7314. The server room table 7310 shows the history of the cooling device power consumption fluctuation type for each server room.
[0039] The server room ID 7311 is an identifier for identifying a server room, and there is no duplication. There is no restriction on the method of assigning IDs. The server room ID 7311 is assigned assuming one ID per room, but even if there are multiple physical rooms, they may be virtually considered as one server room and an ID may be assigned to them.
[0040] The cooling device power consumption fluctuation type 7312 stores whether the fluctuation of the cooling device power consumption is a log that changes continuously or a log that changes in a rectangular shape (discontinuously). The cooling device power consumption is the total power consumption of one or more cooling devices in the server room.
[0041] The start date 7313 stores the time when the cooling device power consumption fluctuation type became that type. The end date 7314 stores the time when the cooling device power consumption fluctuation type became a different type. In the state before changing to a different cooling device power consumption fluctuation type, the end date 7314 that describes the end time is Null.
[0042] 5 is a diagram showing an example of a cooling device power consumption log table 7320. The cooling device power consumption log table 7320 is composed of a plurality of records, and each record has columns of a timestamp 7321, a server room ID 7322, a cooling device ID 7323, and power consumption 7324. The cooling device power consumption log table 7320 is an example of data history of each of a plurality of data sources. One or more cooling devices installed in a server room are an example of an air conditioning system for the server room.
[0043] The data to be stored is not limited to this, and other data may be stored. For example, the set temperature and humidity of the cooling device, the inlet air temperature, the exhaust air temperature, etc. are assumed. In addition, in the case of a water-cooled device, the inlet temperature, the exhaust temperature, etc. are assumed. In addition, although the power consumption of each cooling device is shown in this embodiment, the data may be collectively held in units such as a power panel.
[0044] Timestamp 7321 is an identifier that indicates the date and time that is the subject of the log. In this embodiment, the time unit is indicated as one hour, but this is not limited to this. For example, if logs are kept every minute or every second, that would be the time unit.
[0045] The server room ID 7322 is an identifier for identifying a server room. The same value as the server room ID 7311 included in the server room table 7310 is used to identify the same server room.
[0046] The cooling device ID 7323 is an identifier for identifying a cooling device. There is no duplication within one server room. The same cooling device ID 7323 is assigned to the same cooling device across timestamps 7321. There is no restriction on the method of assigning IDs.
[0047] The power consumption amount 7324 is the power consumption amount of the cooling device of each record at the date and time. In this embodiment, the unit of power consumption is Wh, but this is not limiting.
[0048] 6 is a diagram showing an example of an IT equipment power consumption log table 7330. The server room table 7310 is composed of a plurality of records, and each record has columns of a timestamp 7331, a server room ID 7332, an IT equipment ID 7333, and a power consumption amount 7334. In this embodiment, the power consumption of each computer is shown, but the data may be collectively held in units such as a distribution board.
[0049] The IT device ID 7333 is an identifier for identifying the IT device. Other points are the same as the cooling device ID 7323. The timestamp 7331, server room ID 7332, and power consumption 7334 have the same meanings as the timestamp 7321, server room ID 7322, and power consumption 7324 in the cooling device power consumption log table 7320, respectively, and therefore description thereof will be omitted.
[0050] 7 is a diagram showing an example of a sensor log table 7340. The sensor log table 7340 is made up of a plurality of records, and each record has columns of a timestamp 7341, a server room ID 7342, a sensor ID 7343, and metrics 7344.
[0051] The sensor ID 7343 is an identifier for identifying the sensor. Other aspects are the same as the cooling device ID 7323. In this embodiment, one ID is assigned to each sensor. Furthermore, data is shown in units of one ID, that is, in units of sensors, but this is not limited to this.
[0052] Metrics 7344 is the metrics for the sensor and date and time of each record. In this embodiment, the temperature (Celsius temperature) and humidity (relative humidity) installed in the server room are stored. However, this is not limited to this data, and other data such as Fahrenheit temperature, absolute humidity, air pressure, and air volume may be stored.
[0053] A timestamp 7341 and a server room ID 7342 have the same meaning as the timestamp 7321 and the server room ID 7322 in the cooling device power consumption log table 7320, and therefore a description thereof will be omitted.
[0054] 8 is a diagram showing an example of a weather information log table 7350. The weather information log table 7350 is made up of a plurality of records, and each record has columns of a timestamp 7351 and a metrics 7352. The timestamp 7351 and the metrics 7352 have the same meaning as the timestamp 7341 and the metrics 7344 in the sensor log table 7340, and therefore a description thereof will be omitted.
[0055] In this embodiment, the data included in the metrics 7344 is assumed to be the outside air temperature (Celsius temperature) and humidity (relative humidity) near the data center, but is not limited to this. For example, other data may be held, and expected data include weather, cloud cover, wind volume, wind direction, air pressure, dew point temperature, and sunshine hours. Also, in this embodiment, data for one location is shown, but is not limited to this. In that case, a column including an ID or name for identifying the region may be added to the weather information log table 7350, and it may be indicated in which region each server room is located.
[0056] 9 is a diagram showing an example of a parameter management table 7360. The parameter management table 7360 is composed of a plurality of records, and each record has columns of a server room ID 7361, a parameter ID in the server room 7362, a parameter: threshold value of difference between predicted and actual values 7363, a registration date and time 7364, and an application end date and time 7365. The parameter management table 7360 is an anomaly detection parameter history, and specifically indicates the history of parameters (threshold value of difference between predicted and actual values) for each server room.
[0057] The server room ID 7361 has the same meaning as the server room ID 7322 in the cooling device power consumption log table 7320, so a description thereof will be omitted. The server room parameter ID 7362 is an identifier for identifying parameters when multiple parameters are set for each server room. There is no duplication within one server room. The method of assigning IDs does not matter. Also, in this embodiment, the number of parameters existing for a single server room may change over time. Therefore, the number of server room parameter IDs and records linked to one server room ID may change over time.
[0058] Parameter: threshold value of predicted / actual difference 7363 stores the value of the threshold value of the predicted / actual difference, since the threshold value of the predicted / actual difference is used as a parameter in this embodiment.
[0059] The date and time when the parameter is registered and usage begins is stored in registration date and time 7364. The date and time when usage of the parameter of that record ends is stored in application end date and time 7365. Before another field is stored for the server room ID, the application end date and time column 7365 indicating the date and time when application of the parameter ended is Null.
[0060] 10 is a flowchart 10000 showing data collection and data storage processing executed by the data collection program 6210. In this embodiment, the data collection program 6210 is executed periodically. First, in step S10020, the data collection program 6210 acquires metrics data from any of the managed objects together with the target date and time of the metrics data. In this embodiment, the managed objects are the computer 3000, the cooling device 4000, and the temperature and humidity sensor 5000.
[0061] The data collection program 6210 stores the metrics data acquired in step S10030 in the cooling device power consumption log table 7320, the IT equipment etc. power consumption log table 7330, or the sensor log table 7340. Specifically, in step S10020, if the data collection program 6210 has acquired metrics data of the cooling device 4000, it stores it in the cooling device power consumption log table 7320, if the data collection program 6210 has acquired metrics data of the computer 3000, it stores it in the IT equipment etc. power consumption log table 7330, and if the data collection program 6210 has acquired metrics data of the temperature and humidity sensor 5000, it stores it in the sensor log table 7340.
[0062] In the next step S10040, the data collection program 6210 judges whether or not there is a management target for which data has not been acquired. If the data collection program 6210 judges that there is a management target for which data has not been acquired, the program returns to step S10020, and if it judges that there is no management target for which data has not been acquired, the program executes the subsequent step S10040.
[0063] In the next step S10050, the data collection program 6210 acquires weather data for areas geographically close to the first data center 1000 from sensors installed in the first data center 1000 or from weather information providers, together with the target date and time of the metrics data. In the next step S10060, the data collection program 6210 stores the metrics data acquired in step S10050 in the weather information log table 7350.
[0064] In the next step S10070, the data collection program 6210 determines whether or not there is any unacquired weather data. If the data collection program 6210 determines that there is unacquired weather data, it returns to step S10050, and if it determines that there is no unacquired weather data, it ends the process shown in FIG.
[0065] However, the data acquired and saved by the data collection program 6210 is not limited to this. The type of data to be acquired depends on the type of data to be used for anomaly detection and parameter update. In this case, it is sufficient to acquire the data necessary for anomaly detection and parameter update.
[0066] 11 is a flowchart 11000 showing a cooling device power consumption fluctuation type determination process executed by the cooling device power consumption fluctuation type determination program 6220. In this embodiment, it is executed periodically. For example, it may be executed once a day or once a week. The cooling device power consumption fluctuation type determination program 6220 first acquires the cooling device power consumption within a certain range from the current time from the cooling device power consumption log table 7320 in step S11020. The certain range may be, for example, one week or one month. The cooling device power consumption is the total power consumption of one or more cooling devices in the server room.
[0067] In the next step S11030, it is determined whether the power consumption of a cooling device in a certain server room is a log showing a continuous change or a rectangular change, using the cooling device power consumption data acquired from the cooling device power consumption log table 7320. For example, inverter control of a cooling device may show a power consumption that changes continuously, and ON / OFF control may show a power consumption that changes rectangularly.
[0068] In this embodiment, the kurtosis of the distribution is calculated for the 24-hour moving average of the cooling device power consumption, and if it falls below a certain value, it is considered to be a "continuous change", and if it exceeds a certain value, it is considered to be a "rectangular change". Note that the method of determining the cooling device power consumption fluctuation type and the types of the type are not limited to this. Other statistical values, for example, variance, may be used, and the waveform may be classified into three or more categories. In this way, the cooling device power consumption fluctuation type determination program 6220 determines the type from the characteristics of the cooling device power consumption waveform itself. Waveforms of the same type are considered to be similar.
[0069] In the next step S11040, the cooling device power consumption fluctuation type determination program 6220 acquires from the server room table 7310 the latest cooling device power consumption fluctuation type stored in the table for that server room.
[0070] In the next step S11050, the cooling device power consumption fluctuation type determination program 6220 compares the current cooling device power consumption fluctuation type determined in step S11030 with the latest cooling device power consumption fluctuation type stored in the server room table 7310 obtained in step S11040.
[0071] If the cooling device power consumption fluctuation type acquired in step S11040 is Null, or if the cooling device power consumption fluctuation type determined in step S11030 differs from the cooling device power consumption fluctuation type acquired in step S11040, the process proceeds to step S11060. If not, the process shown in FIG. 11 is terminated.
[0072] When step S11060 is executed, the cooling device power consumption fluctuation type determined in step S11030 and the time (date and time) when that state was reached are stored in the start date 7313 of the server room table 7310. Furthermore, if the end timing corresponding to the cooling device power consumption fluctuation type acquired in step S11040 has not been recorded, the time (date and time) when that state ended is saved in the end date 7314 of the server room table 7310.
[0073] In the case of a state before termination, the end date 7314, which is an item describing the end timing, is Null. Note that the cooling device power consumption fluctuation type 7312 of the server room in question obtained from the server room table is Null if this flow has never been executed, and information is stored otherwise.
[0074] 12 is a flowchart 12000 showing the entire parameter update process executed by the parameter update program 6240, which determines whether or not the parameters need to be updated for the cooling device with a low power consumption, actually updates the parameters, and notifies the user of the result. Inside the parameter update process, the power consumption prediction program 6230 is executed.
[0075] The parameter update program 6240 is a program that is executed periodically for each server room as shown in step S12010. First, in step S12020, the parameter update program 6240 acquires cooling device power consumption data from the cooling device power consumption log table 7320, power consumption data of IT devices such as computers from the IT device power consumption log table 7330, temperature and humidity measured by the sensor 5000 installed in the server room from the sensor log table 7340, and outside air temperature and humidity from the weather information log table 7350, for a certain range from the current time. The certain range may be the same as the certain range in determining the cooling device power consumption fluctuation type, for example.
[0076] In the next step S12030, the power consumption of the cooling device is predicted by the power consumption prediction program 6230 using the data acquired in step S12020. In this embodiment, the parameter update program 6240 is executed every hour, and the power consumption for the next hour is predicted using the data acquired in step S12020. Note that in this embodiment, the sensor log contains two types of data, temperature and humidity, but these are handled by metric type. The method for predicting the power consumption of the cooling device is not limited.
[0077] In the next step S12040, parameters corresponding to the server room and date / time being processed are obtained from parameter management table 7360.
[0078] In the next step S12050, the power consumption of the cooling device predicted in step S12030 is compared with the actual power consumption of the cooling device acquired in step S12020, and the value is compared with the parameter acquired in step S12040 to determine whether the parameter needs to be updated. In this embodiment, the difference between the predicted and actual values is expressed as an absolute value, and the parameter is a threshold value of the difference between the predicted and actual values. However, this is not limited to this. Other examples of the parameters are described in, for example, embodiment 4.
[0079] In this embodiment, the following method is considered for the part where the difference between the forecast and the actual value is compared with the parameters to determine whether the parameters need to be updated: If there is only one parameter at the current time for the server room in the parameter management table 7360, that is, if there is only one parameter ID 7362 within the server room, then that single parameter value is set as the threshold value for the difference between the forecast and the actual value.
[0080] Next, if there are multiple parameters in the server room, i.e. if there are two or more parameters with ID 7362 in the server room, each parameter is compared with the predicted / actual difference to determine whether or not an update is required. In this case, a majority vote is used to determine whether an update is required. For example, if there are five parameters, if the predicted / actual difference exceeds three of the parameters, an update is deemed necessary.
[0081] In the above example, the actual measurement value is the current data of the amount of power consumption of the cooling device, and the update condition is met when the difference between the actual measurement value and the predicted value satisfies any of the above relationships with the threshold value, which is the abnormality detection parameter.
[0082] In the next step S12060, the result of step S12050 is used to determine whether an update is necessary. If it is determined that an update is not necessary, the process shown in Fig. 12 ends. If it is determined that an update is necessary, the process proceeds to step S12070. In step S12070, flow 13000 for calculating similarity between the current and past data of each of the other server rooms and the past data of the server room in question is called in Fig. 13. The similarity is calculated based on the amount of power consumed by the cooling devices. A detailed explanation will be given in the explanation of flow 13000 for calculating similarity between each of the other server rooms.
[0083] In the next step S12080, it is determined whether a similar server room exists using the similarity between the server room and each of the other server rooms and the server room in the past, which is the processing result of step S12070. In this embodiment, a specific value is preset as a threshold value for the similarity, and it is determined whether a similar server room exists based on whether the value is exceeded.
[0084] If a similar server room exists, the server room is extracted as a similar server room, and the process proceeds to step S12090. If a similar server room exists, one or more server rooms are extracted. If no similar server room exists, the process proceeds to step S12120.
[0085] In step S12090, if a similar server room is found, the parameters are updated. In this embodiment, there are roughly two possible parameter update methods: when one similar server room is extracted and when multiple similar server rooms are extracted. When only one similar server room is extracted, the parameter (threshold) of the similar server room is divided by the power consumption of IT equipment, etc. in the similar server room, and the value multiplied by the power consumption of IT equipment, etc. in the server room is used as the new parameter of the server room.
[0086] If multiple similar server rooms are extracted, a method is considered in which the same correction calculation is performed for each similar server room as when only one similar server room is extracted, and all parameters are made new parameters for that server room. In this case, the new parameters for that server room will be the same number as the number of similar server rooms. Based on this, proceed to step S12100.
[0087] In step S12120, the user determines whether or not to perform manual correction based on the results displayed on the user interface (UI). If the user determines not to perform manual correction, the process shown in Fig. 12 ends. The UI used here may be the same as that shown in Fig. 14, but is not limited to this.
[0088] If the user decides to perform manual correction, the process proceeds to step S12130. In step S12130, the user manually inputs new parameters on the UI. Based on this, the process proceeds to step S12100. This does not apply to behavior when no similar server room exists. For example, it is possible not to perform any parameter correction at all.
[0089] In step S12100, the cooling device power consumption fluctuation type determination program 6220 saves the new parameters output in step S12090 in the parameter management table 7360. When saved, the application end date and time 7365 becomes Null. If there is already a record in the parameter management table 7360 for the server room that is the target of this flow and its application end date and time 7365 is Null, it is changed to the same date and time as the registration date and time 7364 of the added record.
[0090] In the next step S12110, the parameter update is notified to the user via the UI, an embodiment of which will be described later with reference to Fig. 14. This ends the processing of Fig. 12.
[0091] FIG. 13 is a flowchart 13000 showing the contents of the inter-server room similarity calculation process S12070 executed by the cooling device power consumption fluctuation type determination program 6220 in the parameter update flow 12000.
[0092] The cooling device power consumption fluctuation type determination program 6220 first obtains, in step S13020, a cooling device power consumption fluctuation type at the current time of the server room determined in advance from the server room table 7310, and further extracts the period of server rooms having the same fluctuation type as the server room, including the current and past data of other server rooms and the past period of the server room. This increases the possibility of obtaining a similar period. As another example, only the current information of other server rooms may be referenced.
[0093] At this time, periods in which the fluctuation type is Null are ignored. Also, if the cooling device power consumption fluctuation type in the server room table 7310 for that server room is Null, the parameter update program 6240 is not executed in the first place.
[0094] In the next step S12030, it is determined whether or not a period of time with the same cooling device power consumption fluctuation type exists for the server room in step S12020. If so, the process proceeds to step S13040 in which the similarity with periods of the same cooling device power consumption fluctuation type is calculated, and if not, the process proceeds to step S13060.
[0095] If step S12030 finds that there is a period of time for the server room with the same cooling device power consumption fluctuation type, then in step S12040, cooling device power consumption data for a specified range from the current time is obtained from cooling device power consumption log table 7320, power consumption data of IT equipment such as computers from IT equipment power consumption log table 7330, and outside temperature, humidity, etc. from weather information log table 7350.
[0096] In the next step S12050, the similarity for a period having the same fluctuation type as the server room is calculated using the data acquired in step S12040. The method of calculating the similarity is not limited, but one example is to determine it based on the correlation with external metrics that affect the power consumption of the cooling device. In this embodiment, the similarity is calculated based on the impact of the power consumption of IT equipment, etc. on the power consumption of the cooling device and the impact of the outside air temperature on the power consumption of the cooling device.
[0097] Specifically, the correlation coefficient between the power consumption of the cooling device and the power consumption of the IT devices and the outside temperature is calculated, and the root mean square error is calculated for each correlation coefficient for a period of a server room with the same fluctuation type as the server room, and the sum of these is regarded as the similarity. In other words, the sum of the root mean square error of the correlation coefficient between the power consumption of the cooling device and the power consumption of the IT devices and the correlation coefficient between the power consumption of the cooling device and the outside temperature of the IT device is regarded as the similarity. This ends the processing of FIG. 13. Note that correlation with other metrics may be used, or only one of the above two metrics may be used. By using the correlation between the power consumption of the cooling device and the metrics, data with similar characteristics can be more appropriately extracted.
[0098] In step S13060, if it is determined in step S13030 that there is no period of the server room with the same fluctuation type, the similarity between the server room and all server rooms is set to the lowest value. In this embodiment, it is set to 0. This ends the processing of FIG. 13. By referring to both the fluctuation type and the similarity, data similar to the target data can be more appropriately extracted. As described above, the similarity between periods of the cooling device power consumption of the server room is determined by a preset method. Note that the calculation of the similarity of the change in the cooling device power consumption may be determined by other methods. The fluctuation type may not be referenced to search for a similar period. The calculation of the similarity may be omitted and the similarity of periods of the same fluctuation type may be set to a value exceeding the similarity threshold.
[0099] 14 is a screen output diagram showing the calculation results of the parameter update program 6240. This screen output is implemented in step S12110 of the parameter update flow 12000. This screen output displays the server room for which the parameter update was implemented and information about the server room, information about server rooms similar to that server room, the reason why the server room was determined to be similar, and the reason why new parameters were calculated.
[0100] The upper part of the output screen 6250A shows the server room for which the parameter update was performed in 6250A1 and information about the server room. Specifically, the server room for which the parameter update was performed is shown in a "Server Room" column, the parameter update status is shown in a "Status" column, similar server rooms extracted in step S12080 of Fig. 12 are shown in a "Similar Server Rooms" column, the cooling device power consumption fluctuation type of the server room at the current time acquired in step S13020 of Fig. 13 is shown in a "Cooling Device Power Consumption Fluctuation Type" column, and the power consumption of IT devices, etc. of the server room acquired in step S12020 of Fig. 12 is shown in a "Power Consumption of IT Devices, etc." column.
[0101] In the time series diagram indicated by the reference numeral 6250A2, the estimated values obtained in step S12030 in Fig. 12 are indicated by dotted lines, and the actual measured values acquired in S12020 are indicated by solid lines. The predicted values and the actual measured values diverged midway, and this time series diagram shows a case in which it was determined in step S12060 that parameter updating was necessary.
[0102] Reference numeral 6250A3 displays the result of the similarity between the server room similar to the server room calculated in step S13050 of Fig. 13. Specifically, the names of the server rooms determined to have the same variation type in step S13020 of Fig. 13 are entered in a "Server Room" column, the similarity between the server rooms calculated in step S13050 is entered in a "Similarity" column, which period of data was applied is entered in a "Data Period" column, the parameters of each server room in the data period acquired in S12040 of Fig. 12 are entered in a "Parameters" column, and the power consumption of IT devices, etc. in that period is entered in a "Power Consumption Amount of IT Devices, etc." column.
[0103] Reference numeral 6250A4 notifies the user of a sentence to help him understand the reason why the parameter update notification occurred and the processing content.
[0104] The button 6250A5 shown at the bottom of Fig. 14 is a button for transitioning to an example of a display screen of the parameter update result on the second page shown in Fig. 15. Also, the button 6250A6 is a button for ending viewing of the notification.
[0105] Fig. 15 is a diagram showing an example of a display screen of the parameter update results on the second page transitioned to from Fig. 14. This screen output displays the data acquired in step S13040 of Fig. 13 in a time series diagram indicated by reference numeral 6250B1, and displays the similarity calculated in step S13050 and values calculated during that process at reference numeral 6250B6.
[0106] The time series diagram indicated by reference numeral 6250B1 displays data for that period for the server room in question and for server rooms with the same fluctuation type. In this embodiment, the amount of power consumed by cooling devices is shown here, but the metrics displayed can be changed using buttons indicated by reference numerals 6250B2, 6250B3, 6250B4, and 6250B5. The metrics to be displayed are the metrics used in the similarity calculation in step S13050, and the metrics actually displayed are not limited to those given in this embodiment. In addition, the button indicated by reference numeral 6250B6 may be used to display metrics for server rooms determined not to be similar in step S12080 of FIG. 12.
[0107] The table with the reference numeral 6250B7 shows the similarity between the server room and the server room having the same fluctuation type and the value used to calculate it. Specifically, the names of the server room and the server rooms determined to have the same fluctuation type are entered in the "Server Room" column, and the similarity between the server room and the server rooms determined to have the same fluctuation type calculated in step S13050 is entered in the "Similarity" column.
[0108] In this embodiment, the similarity is defined as the sum of the root mean square error calculated for each of "correlation between air conditioning power consumption and power consumption of IT devices, etc." and "correlation between air conditioning power consumption and outdoor temperature." The definition is shown in the field indicated by reference numeral 6250B8. Therefore, the two calculation results are entered in the columns "correlation between air conditioning power consumption and power consumption of IT devices, etc." and "correlation between air conditioning power consumption and outdoor temperature." The types of values shown in the table indicated by reference numeral 6250B7 may be indices used when calculating the similarity, and are not limited to this.
[0109] In this embodiment, even if the features change over time, other parameters that maintain a similar format are utilized, and the parameters are updated based on similar data to adapt to the changed environment, making it possible to accurately detect inefficiencies in power consumption. This point is the same as in the other embodiments. EXAMPLES
[0110] This embodiment shows another embodiment of step S12090 for updating parameters when a similar server room exists in embodiment 1. With reference to Figs. 1 to 15, the following describes the parts that have been changed from embodiment 1.
[0111] In step S12090 in Fig. 12, if a similar server room period exists, the parameters are updated. In addition to the method shown in the first embodiment, the following parameter update methods are possible. If one similar server room period is extracted, there is a method of using the parameters of the similar server room period as is. Also, if multiple similar server room periods are extracted, the following three update methods are possible as embodiments.
[0112] In the first method, all the parameters of each similar server room are used as they are, in the same way as when one similar server room is extracted in the first embodiment. In this case, the new parameters of the server room will be the same number as the number of similar server rooms. The decision on whether to update can be made by majority vote.
[0113] As a second method, for each similar server room, the average of the parameters of each server room weighted by the power consumption of IT equipment in the server room is calculated, that is, the weighted average is calculated, and the result is used as the parameter. At this time, the new parameter for the server room becomes one.
[0114] In the third method, the average value of the parameters of each similar server room is set as the parameter. At this time, the new parameter of the server room becomes one. Based on this, proceed to step S12100. Thus, although the second and third methods use the statistical value of the parameter, other statistical values may be used. EXAMPLES
[0115] The present embodiment shows another embodiment of step S13050 for calculating the similarity with each server room in the flow 13000 for calculating the similarity with each other server room in the first embodiment.
[0116] With reference to Figs. 1 to 15, the following describes the changes from the first embodiment. In step S13050 in Fig. 13, the correlation coefficient between the power consumption of the cooling device and the power consumption of IT devices, etc., and the outside temperature is calculated, and the root mean square error is calculated for each correlation coefficient for a period of a server room with the same fluctuation type as the server room, and the sum of these is used as the similarity. Instead, the partial regression coefficient is calculated when multiple regression analysis is performed using the same data, with the power consumption of the cooling device set as the objective variable and the power consumption of IT devices, etc., and the outside temperature set as the explanatory variables. There is a method of using the root mean square error as the similarity for the partial regression coefficient of a server room with the same fluctuation type as the server room.
[0117] As another example, similar data is visualized in the form of a graph in the form of an image, which is then input into a convolutional neural network to extract features, and the Euclidean distance or cosine similarity is used as the similarity measure for each of the resulting vectors.
[0118] Here, the power consumption of IT devices and the outside temperature are taken as examples, but this is not limiting. For example, a method using the outside humidity or the room temperature is conceivable. Furthermore, data required for calculating this similarity can be acquired in step S13040. For example, the temperature and humidity measured by the sensor 5000 installed in the server room are acquired from the sensor log table 7340. Note that in this embodiment, the sensor log contains two types of data, temperature and humidity, but they are handled by metric type. EXAMPLES
[0119] The present embodiment shows another embodiment of steps S12030, S12040, and S12050 relating to the determination of whether or not a parameter update is required in the parameter update flow 12000 in the first embodiment.
[0120] With reference to Figs. 1 to 15, the following will show the changes from the first embodiment. In the determination of whether the parameter needs to be updated in steps S12030, S12040, and S12050 in Fig. 12, the determination is made based on whether the predicted / actual difference exceeds a threshold in the first embodiment, but other methods may be used. For example, the model 6330 for detecting an abnormality in the power consumption of the cooling device is used as a rule-based determination criterion. In step S12030, the data acquired in step S12020 is used to calculate a value for making a rule-based determination regarding the fluctuation of the log or the metrics calculated from the log. For example, a method is conceivable in which an abnormality is detected when the rate of increase in the total power consumption of the cooling devices in a server room during a certain period exceeds the value of a determination parameter.
[0121] In the next step S12040, a criterion parameter for determining whether or not an update is required based on a rule base is acquired. In the next step S12050, whether or not the parameter needs to be updated is determined based on the parameter and the calculated value. However, the determination of whether or not an update is required is not limited to this. For example, the log may use the rate of change in indoor temperature or the rate of change in indoor humidity within a certain range from the current time, or the variance value and the rate of change of the amount of power consumed by cooling devices in the same server room. The certain range may be, for example, one week or one month, and may be the same as the certain range in determining the type of fluctuation in the amount of power consumed by cooling devices. EXAMPLES
[0122] In the first embodiment, the parameter to be updated is the parameter used as the threshold value for anomaly detection, which is the predicted / actual difference threshold value 7363, and is described in the parameter management table 7360. However, another parameter may be updated. For example, a parameter may be updated when another anomaly detection is performed simultaneously.
[0123] In the above embodiment, the predicted / actual difference threshold, which is an anomaly detection parameter, is referenced for both notifying the user of an anomaly in the power consumption of the cooling device and updating the parameters. In another example, the anomaly detection parameter may be used for updating the parameters, and the user notification may be omitted. For example, the anomaly detection parameter is set as a first anomaly detection parameter, and a second anomaly detection parameter of a different type is further set.
[0124] For example, a second abnormality detection parameter is set for cooling device power consumption, and the system notifies a user with an alert when the cooling device power consumption exceeds the second abnormality detection parameter (threshold value). The system updates the second abnormality detection parameter together with updating the first abnormality detection parameter. The relationship between the second abnormality parameter and the first abnormality detection parameter may be defined by a function, a lookup table, or the like. The second abnormality detection parameter may be data from a data source different from that of the cooling device power consumption.
[0125] The above-mentioned embodiments can be applied to data obtained from devices or systems other than an air conditioning system including a cooling device or other air conditioning devices. In the above-mentioned embodiments and modifications, the functional block configurations are merely examples. Some functional configurations shown as separate functional blocks may be integrated, or a configuration shown in one functional block diagram may be divided into two or more functions. Also, some of the functions of each functional block may be provided by other functional blocks.
[0126] In each of the above-described embodiments and modifications, the management computer 6000 may be provided with an input / output interface (not shown), and a program may be loaded from another device when necessary via the input / output interface and a medium available to the management computer 6000. Here, the medium refers to, for example, a storage medium that is detachable from the input / output interface, or a communication medium, that is, a wired, wireless, optical, or other network, or a carrier wave or digital signal that propagates through the network. In addition, some or all of the functions realized by the program may be realized by a hardware circuit or an FPGA.
[0127] The present invention is not limited to the above-mentioned embodiment, but includes various modified examples. For example, the above-mentioned embodiment has been described in detail to clearly explain the present invention, and is not necessarily limited to those including all of the configurations described. It is also possible to replace a part of the configuration of one embodiment with the configuration of another embodiment, and it is also possible to add the configuration of another embodiment to the configuration of one embodiment. It is also possible to add, delete, or replace a part of the configuration of each embodiment with another configuration.
[0128] In addition, the above-mentioned configurations, functions, processing units, etc. may be realized in part or in whole by hardware, for example, by designing them as integrated circuits. In addition, the above-mentioned configurations, functions, etc. may be realized in software by a processor interpreting and executing a program that realizes each function. Information such as the program, table, file, etc. that realizes each function can be stored in a memory, a recording device such as a hard disk or SSD (Solid State Drive), or a recording medium such as an IC card or SD card.
[0129] In addition, the control lines and information lines shown are those that are considered necessary for the explanation, and not all control lines and information lines in the product are necessarily shown. In reality, it can be considered that almost all components are connected to each other. [Explanation of symbols]
[0130] 1000: Data Center 3000: Calculator 4000: Cooling device 5000: Temperature and humidity sensor 6000: Management computer 6220: Cooling device power consumption fluctuation type determination program 6230: Power consumption prediction program 6240: Parameter update program 6250: Example of parameter update result display screen 7000: Datastore 7310: Server Room Table 7320: Cooling unit power consumption log table 7330: Power consumption log table for IT devices, etc. 7340: Sensor log table 7350: Weather information log table 7360: Parameter management table
Claims
1. 1. An anomaly detection system, comprising: one or more processors; one or more storage devices; The one or more storage devices include Data history for each of the multiple data sources, anomaly detection parameter history indicating a history of anomaly detection parameters for detecting anomalies in data from each of the multiple data sources; The one or more processors: Obtain current data of a first data source of the plurality of data sources; When a relationship between the current data and the current anomaly detection parameter of the first data source satisfies a predetermined update condition, extracting one or more periods determined to be similar to a period including the current data by a preset method from at least a portion of the data history of the multiple data sources; obtaining anomaly detection parameters for the one or more time periods from anomaly detection parameter histories for at least a portion of the plurality of data sources; An anomaly detection system that updates the current anomaly detection parameters based on the acquired anomaly detection parameters for the one or more periods.
2. The anomaly detection system according to claim 1 , The data history of the multiple data sources indicates the power consumption history of multiple air conditioning systems.
3. The anomaly detection system according to claim 2, The one or more storage devices store a prediction model that includes, as an input, the past power consumption of the target air conditioning system and predicts the future power consumption of the target air conditioning system; the anomaly detection parameter is a threshold associated with a difference between a predicted value and an actual value of power consumption of the air conditioning system; The one or more processors: inputting data including historical power consumption of the first data source into the prediction model to obtain a prediction of future power consumption of the first data source; Detecting an anomaly in the power consumption of the first data source based on a comparison result between the difference between the predicted value and the actual measured value of the power consumption of the first data source and a current threshold value; The anomaly detection system updates the current threshold based on the threshold for the one or more time periods.
4. The anomaly detection system according to claim 3, An anomaly detection system, wherein the input of the predictive model includes the outdoor temperature of the target air conditioning system.
5. The anomaly detection system according to claim 1 , The one or more processors determine a statistical value of the anomaly detection parameter for the one or more time periods as an anomaly detection parameter after updating the current anomaly detection parameter.
6. The anomaly detection system according to claim 1 , The one or more processors: determining the update condition by majority voting using one or more current anomaly detection parameters of the first data source; The anomaly detection system determines all of the anomaly detection parameters for the one or more time periods to be the updated anomaly detection parameters of the first data source.
7. The anomaly detection system according to claim 2, The anomaly detection system, wherein the determination of the similarity of the period is based on a similarity between waveforms of the power consumption of the air conditioning system and a similarity of correlation between the power consumption and external metrics that affect the power consumption.
8. The anomaly detection system according to claim 1 , A period similar to the period including the current data is searched for in the data history of the first data source and another data source different from the first data source.
9. The anomaly detection system according to claim 1 , An anomaly detection system in which, when the one or more processors update the current anomaly detection parameters, they present to a user current and past data from the first data source, information on the updated anomaly detection parameters, and information on the anomaly detection parameters for the one or more periods.
10. 1. A method performed by an anomaly detection system, comprising: The anomaly detection system includes: Data history for each of the multiple data sources, anomaly detection parameter history indicating a history of anomaly detection parameters for detecting anomalies in data from each of the multiple data sources; The method comprises: The anomaly detection system obtains current data of a first data source in the plurality of data sources; When a relationship between the current data and the current anomaly detection parameter of the first data source satisfies a predetermined update condition, the anomaly detection system extracts one or more periods determined to be similar to a period including the current data by a preset method from data histories of at least a portion of the multiple data sources; the anomaly detection system deriving anomaly detection parameters for the one or more time periods from anomaly detection parameter histories for at least a portion of each of the plurality of data sources; The method, wherein the anomaly detection system updates the current anomaly detection parameters based on the acquired anomaly detection parameters for the one or more time periods.
Citation Information
Patent Citations
Image grid with selectively prominent images
US10896478B2
Cited By
Method and device for determining detection frequency of OLT (Optical Line Terminal) equipment and electronic equipment
CN121099224A
Anomaly detection method, and device, storage medium and program product
WO2026153142A1
