JP7166395B1 - MONITORING SYSTEM, MONITORING METHOD, AND MONITORING PROGRAM - Google Patents

Abstract

A monitoring system, a monitoring method, and a program for monitoring a target for abnormality determination are provided. Kind Code: A1 A monitoring system includes one or more sensors provided on a target for determining an abnormality, and sensor data detected by each of the one or more sensors provided on the target for determining whether or not there is an abnormality. 1 processing unit, and a second processing unit that determines whether or not to issue an abnormality report based on sensor data detected by each of the plurality of sensors when the first processing unit determines that there is an abnormality , the second processing unit learns the relationship between the sensor data detected by the sensors provided for each of the plurality of targets and the abnormality related to the target, and uses the learning result to determine whether the first processing unit is abnormal. It is determined whether or not to report an abnormality for a target having a sensor that outputs sensor data determined to be present. [Selection drawing] Fig. 1

Description

Application of Article 30, Paragraph 2 of the Patent Act (1) Date November 13, 2020 (Duration: November 12 to November 13) Meeting Name, Venue IM Co., Ltd. / Seiko Solutions Co., Ltd. / Cosmo Co., Ltd. Sponsored by Society 5.0 for SDGs WEEK COMPUS 2020 Online Online event using Internet communication (Zoom) <Reference> Overview of COMPUS 2020 <Reference> COMPUS 2020 presentation materials

Application of Article 30, Paragraph 2 of the Patent Act (2) Date January 20, 2021 Meeting name, venue Hosted by the Digital Business Innovation Center (DBIC) Enterprise Transformation Practice Series New Year Special Challenger talks about tomorrow” Held online using Internet communication (Zoom) <Document> Overview of Web Seminar Corporate Transformation Practice Series <Document> Web Seminar Corporate Transformation Practice Series presentation materials

Application of Article 30, Paragraph 2 of the Patent Law (3) Date of posting on the website April 1, 2021 Website address https://www. iim. co. jp/https://www. iim. co. jp/case/t-hoan/ https://www. iim. co. jp/products/luina/ https://www. iim. co. jp/products/luina/? utm_source=20210405&utm_medium=email <Reference> “LUiNa” Product Description Webpage

Application of Article 30, Paragraph 2 of the Patent Act (4) Website publication date April 5, 2021 Website address [PR-related/posting website address] https://www. iim. co. jp/downloadfiles/luina_case_tohoku. pdfhttps://www. seiko-sol. co. jp/case/https://www. iim. co. jp/case/t-hoan/https://www. iim. co. jp/case/https://www. iim. co. jp/https://form. seiko-sol. co. jp/m/iim_dl_luinahttps://www. iim. co. jp/downloadfiles/luina_case_tohoku. pdfhttps://www. seiko-sol. co. jp/case/ <Material> “LUiNa” introduction article example web page

Application of Article 30, Paragraph 2 of the Patent Act (5) Date April 14-16, 2021 Meeting name, venue Interop Tokyo 2021 Makuhari Messe (International Exhibition Hall) sponsored by the Interop Tokyo Executive Committee <Document> Interop Tokyo 2021 Exhibit Guide < Materials > Overview of Interop Tokyo 2021

The present invention relates to a monitoring system, a monitoring method, and a monitoring program.

The chief electrical engineer system is a system to ensure the safety and security of electrical safety, but the system will be maintained stably in the future due to the aging of chief electrical engineers and the decrease in the number of new employees to outsourced contractors. is in a difficult situation. In addition, even though the flow of people has been restrained due to the recent corona disaster, ensuring security is emphasized, especially the maintenance and operation of industrial infrastructure, medical institutions and nursing homes.

Under such circumstances, while the public and private sectors are working together to secure chief electrical engineers, IoT (Internet of Things) and AI (artificial intelligence) have been introduced to ensure safety even with a small number of personnel, improving safety and work efficiency. A “smart safety” system is being considered.
As for the remote system, there is known a technique for easily executing the output control of the power conditioner (see, for example, Patent Literature 1).
90% of private electrical equipment is small and medium-sized enterprises using the outsourcing system. Most of them use the outsourcing system to ensure electrical safety. Although some of them operate low-voltage insulation monitoring devices and play a role in electrical safety, the system for collecting data remotely is rarely used.

JP 2020-182259 A

Regular inspections are legally required for power receiving and transforming equipment for small and medium-sized private electric facilities. Currently, engineers such as chief electrical engineers visit the site and conduct inspections using all five senses, so ensuring safety and security depends on the competence of the engineers.
Signs of failure, such as unusual noises and odors that accompany deterioration of equipment over time, can be detected only when an engineer is inspecting the equipment. However, at the time of inspection, there is a risk that a sign of failure may be overlooked because it is not always the case.
In addition, due to the recent situation, it is becoming difficult to secure human resources for engineers and dispatch them to the site for inspection, and there is a demand for more efficient inspection work.
In addition, due to the aging of skilled engineers and the difficulty in securing new personnel, it is becoming difficult to pass on technology.
In addition, such problems are not limited to inspections of small and medium-sized electric facilities for private use. For example, in plants and factories, workers periodically inspect devices. Also, in building security companies, inspectors regularly monitor the inside and outside of the building. Such inspection and monitoring work by humans poses the same problem of securing human resources as in inspections of small and medium-sized electric facilities for private use, and at the same time, efficiency is required.
SUMMARY OF THE INVENTION It is an object of the present invention to provide a monitoring system, a monitoring method, and a program for monitoring an object for abnormality determination.

(1) In view of the above-described problems, a monitoring system according to an aspect of the present invention is a monitoring system that monitors any one of a power receiving and transforming facility, a plant, a factory, and a building, and monitors a target for abnormality determination. Therefore, a plurality of sensors provided on the target and time-series data of sensor data constantly detected by each of the plurality of sensors provided on the target are checked for each of the sensors to determine whether they behave differently than normal. and when it is determined that the time-series data of some of the sensor data out of the plurality of sensor data is moving differently than in normal times, the part of the sensor data determined to be moving differently than in normal times a first processing unit that identifies the sensor data and acquires the timing at which the sensor data is acquired ; When it is determined that the sensor is moving, acquiring the timing at which the part of the sensor data determined to be moving differently from normal is acquired, and based on the acquired timing , each of the plurality of sensors a second processing unit that acquires the sensor data detected by and determines whether or not to issue an anomaly alarm based on the acquired plurality of sensor data, the second processing unit includes: The relationship between the sensor data detected by each of the plurality of sensors provided in each of the sensors and the abnormality related to the object is learned, and using the learning result, the first processing unit detects a movement different from normal It is a monitoring system for determining whether or not to issue an anomaly report for a target equipped with a sensor that has output the part of sensor data determined to be present.
(2) In the monitoring system according to an aspect of the present invention, the first processing unit is based on time-series data of sensor data detected by each of the plurality of sensors provided for each of the plurality of targets, For sensor data corresponding to each of the plurality of sensors, unsupervised learning of patterns between sensor data is performed, and using the learning result, the sensor data detected by each of the plurality of sensors is already It is determined whether or not the time-series data corresponding to each of the plurality of acquired sensors can be classified, and if the classification is not possible, it is determined that the movement is different from normal.
(3) A monitoring system according to an aspect of the present invention is a monitoring system that monitors any one of a power receiving and transforming facility, a plant, a factory, and a building. and the time-series data of the sensor data constantly detected by each of the plurality of sensors provided in the object and the plurality of sensors provided in the object. When it is determined that the time-series data of a part of the sensor data out of the sensor data is moving differently than in normal times, the part of the sensor data determined to be moving differently than in normal times is specified. a first processing unit that acquires the timing at which the sensor data is acquired; and a movement in which the time-series data of a part of the plurality of sensor data provided to the target by the first processing unit is different from normal time. when it is determined that the sensor is moving differently than normal, acquiring the timing at which the part of the sensor data determined to be moving differently from normal is acquired, and based on the acquired timing, each of the plurality of sensors a second processing unit that acquires the detected sensor data and determines whether or not to issue an abnormality alarm based on the sensor data that is detected by each of the plurality of acquired sensors; Based on time-series data of sensor data detected by each of the plurality of sensors provided for each of the plurality of targets, unsupervised detection of patterns between sensor data for sensor data corresponding to each of the plurality of sensors. learning, and using the result of the learning, whether the sensor data detected by each of the plurality of sensors can be classified into time-series data corresponding to each of the plurality of sensors already acquired If it cannot be classified, it is determined that the object is moving differently from normal, and the second processing unit combines sensor data detected by each of the plurality of sensors provided for each of the plurality of targets with The sensor that has learned the relationship with the abnormality related to the object and has output the part of the sensor data determined that the first processing unit is moving differently from normal using the result of the learning It is a monitoring system that determines whether or not to issue an anomaly report for a provided target.
(4) In the monitoring system according to an aspect of the present invention, the first processing unit learns the sensor data detected by each of the plurality of sensors, and then re-learns by using most recent sensor data preferentially. do.
(5) In the monitoring system according to an aspect of the present invention, the second processing unit selects the part of the sensor that detected the part of the sensor data determined to be moving differently than normal Of the sensor data detected by each of the plurality of sensors provided on the object, sensors other than the sensor that detected the sensor data determined to be moving differently than normal Based on the sensor data detected by the sensor, it is determined whether or not to report an abnormality.
(6) In the monitoring system according to an aspect of the present invention, the first processing unit extracts a plurality of sensor data based on sensor data detected by each of the plurality of sensors provided on the object. It learns the relationship between each and the environment in which the object whose anomaly is to be determined is provided.
(7) In the monitoring system according to an aspect of the present invention, the second processing unit may detect an abnormality determined based on sensor data detected by each of the plurality of sensors provided for each of the plurality of targets. update the relationship between each of the plurality of sensor data and the abnormality of each of the plurality of devices included in the abnormality determination target based on the state of each of the plurality of devices included in the determination target .
(8) In the monitoring system according to one aspect of the present invention , the second processing unit may perform multiple It learns the relationship between the combination of the number of sensor data and the abnormality of each of the plurality of devices included in the object of abnormality determination.
(9) In the monitoring system according to an aspect of the present invention, the second processing unit may perform multiple The relationship between the combination of the number of sensor data and the failure level of each of the plurality of devices included in the object for determining abnormality is learned.
(10) In the monitoring system according to an aspect of the present invention, the second processing unit may detect an abnormality determined based on sensor data detected by each of the plurality of sensors provided for each of the plurality of targets. Based on the failure level of each of the plurality of devices included in the target for determining the abnormality, the relationship between the combination of the plurality of sensor data and the failure level of each of the plurality of devices included in the target for determining abnormality Update.
(11) In the monitoring system according to one aspect of the present invention, each of the one or more sensors is an ultrasonic sensor, an ozone sensor, an odor sensor, or a sound wave sensor.
(12) The monitoring system according to an aspect of the present invention further includes one or more second sensors other than the plurality of sensors provided on the object for determining abnormality, wherein the first processing unit Alternatively, it is determined whether the sensor data detected by each of the plurality of second sensors is abnormal.

(13) A monitoring method according to an aspect of the present invention monitors an object for determining an abnormality. a step of determining whether or not each sensor is moving, and when it is determined in the determining step that the time-series data of some of the plurality of sensor data are moving differently than normal a step of identifying the part of the sensor data determined to be moving differently than normal and obtaining the timing of obtaining the sensor data; acquiring sensor data detected by each of the plurality of sensors based on the acquired sensor data; In the step of determining whether a motion different from normal is performed using a learning result of a relationship between sensor data detected by each of the plurality of sensors provided for each of the plurality of targets and abnormalities related to the plurality of targets monitoring any of the receiving and transforming equipment, the plant, the factory, and the building, for determining whether or not to issue an anomaly report for the target equipped with the sensor that outputs the part of the sensor data determined to be It is a monitoring method executed by a computer of a monitoring system that
(14) A monitoring method according to an aspect of the present invention monitors an object for determining an abnormality, so that time-series data of sensor data constantly detected by each of a plurality of sensors provided on the object exhibits movement different from normal time. a step of determining for each of the sensors whether or not the time-series data of a part of the plurality of sensor data provided for the object in the determining step behaves differently than in normal times; a step of identifying the part of the sensor data determined to be moving differently from normal when it is determined that the part of the sensor data is moving, and acquiring the timing of acquiring the sensor data; acquiring sensor data detected by each of the plurality of sensors based on the timing at which the In the step of determining for each sensor, based on time-series data of sensor data detected by each of the plurality of sensors provided for each of the plurality of targets, sensor data corresponding to each of the plurality of sensors, Unsupervised learning is performed on patterns between sensor data, and using the learning result, the sensor data detected by each of the plurality of sensors corresponds to each of the plurality of sensors that have already been acquired. In the step of determining whether or not the object can be classified into time-series data, determining that the object is moving differently than normal when the classification is not possible, and determining whether or not to issue an abnormality alarm, each of the plurality of objects The relationship between the sensor data detected by each of the plurality of sensors provided and the abnormality related to the object is learned, and the learning result is used to determine whether the motion is different from normal in the step of determining for each of the sensors. Monitoring for monitoring any one of the receiving and transforming equipment, the plant, the factory, and the building, for determining whether or not to issue an anomaly report for the target equipped with the sensor that output the part of the sensor data determined to be It is a monitoring method performed by the computer of the system.

(15) A monitoring program according to an aspect of the present invention causes a computer of a monitoring system that monitors any one of a power receiving and transforming facility, a plant, a factory, and a building to monitor a target for determining an abnormality. a step of determining for each sensor whether or not time-series data of sensor data constantly detected by each of a plurality of sensors detected by the sensor is moving differently from normal times; When it is determined that the time-series data of a part of the sensor data is moving differently than in normal times, the part of the sensor data determined to be moving differently than in normal times is specified, and the sensor data acquiring sensor data detected by each of the plurality of sensors based on the timing of acquiring the part of the sensor data, and based on the acquired plurality of sensor data and a step of determining whether or not to issue an anomaly alarm, and in the step of determining whether to issue an anomaly alarm, each of the plurality of sensors provided for each of the plurality of targets detects Using the learning result of the relationship between the sensor data and the abnormality related to the target, an abnormality notification is issued for the target equipped with the sensor that outputs the part of the sensor data that is determined to be moving differently than normal. It is a monitoring program that determines whether or not
(16) A monitoring program according to an aspect of the present invention causes a computer of a monitoring system that monitors any one of a power receiving and transforming facility, a plant, a factory, and a building to monitor a target for determining an abnormality. a step of determining for each sensor whether or not time-series data of sensor data constantly detected by each of a plurality of sensors detected by each of the sensors is moving differently from normal times ; When it is determined that the time-series data of a part of the sensor data is moving differently than in normal times, the part of the sensor data determined to be moving differently than in normal times is specified. and acquiring the timing of acquiring the sensor data;
Acquiring the sensor data detected by each of the plurality of sensors based on the timing of acquiring the part of the sensor data, and determining whether or not to issue an abnormality alarm based on the acquired plurality of sensor data and the step of determining for each of the sensors, based on the time-series data of the sensor data detected by each of the plurality of sensors provided for each of the plurality of targets, the determination of the plurality of sensors Unsupervised learning of patterns between sensor data is performed for sensor data corresponding to each, and using the learning result, the sensor data detected by each of the plurality of sensors is a plurality of already acquired sensor data In the step of determining whether or not the sensor can be classified into time-series data corresponding to each of the sensors, determining that the sensor is moving differently from normal when the classification cannot be performed, and determining whether or not an abnormality alarm is to be issued , learning the relationship between sensor data detected by each of the plurality of sensors provided for each of the plurality of targets and an abnormality related to the target, and using the learning result to determine for each of the sensors A monitoring program for determining whether or not to issue an anomaly report for a target equipped with a sensor that outputs part of the sensor data that is determined to be moving differently from normal in the step of performing.

According to the present invention, it is possible to provide a monitoring system, a monitoring method, and a program capable of monitoring a target for abnormality determination.

BRIEF DESCRIPTION OF THE DRAWINGS It is a figure which shows an example of the outline of the monitoring system which concerns on embodiment of this invention. It is a figure which shows an example of operation|movement of the monitoring apparatus contained in the monitoring system which concerns on this embodiment. It is a figure which shows an example of the sensor server and monitoring apparatus which are contained in the monitoring system which concerns on this embodiment. It is a figure which shows an example of operation|movement of the monitoring system which concerns on this embodiment. It is a figure which shows an example of operation|movement of the monitoring apparatus contained in the monitoring system which concerns on the modified example 1 of embodiment. It is a figure which shows an example of the monitoring apparatus contained in the monitoring system which concerns on the modified example 1 of embodiment. 9 is a flow chart showing an example of the operation of the monitoring system according to Modification 1 of the embodiment. It is a figure which shows an example of the outline of the monitoring system which concerns on the modification 2 of embodiment. It is a figure which shows an example of the monitoring apparatus contained in the monitoring system based on the modified example 2 of embodiment. 9 is a flow chart showing an example of the operation of the monitoring system according to Modification 2 of the embodiment. It is a figure which shows an example of the monitoring apparatus contained in the monitoring system based on the modified example 3 of embodiment. It is a flow chart which shows an example of operation of a surveillance system concerning modification 3 of this embodiment.

Next, the monitoring system, monitoring method, and monitoring program of this embodiment will be described with reference to the drawings. The embodiments described below are merely examples, and embodiments to which the present invention is applied are not limited to the following embodiments.
In addition, "based on XX" in the present application means "based on at least XX", and includes cases based on other elements in addition to XX. Moreover, "based on XX" is not limited to the case of using XX directly, but also includes the case of being based on what has been calculated or processed with respect to XX. "XX" is an arbitrary element (for example, arbitrary information).

(embodiment)
BEST MODE FOR CARRYING OUT THE INVENTION Hereinafter, embodiments of the present invention will be described with reference to the drawings. Configurations having the same or similar functions are denoted by the same reference numerals, and overlapping descriptions of the configurations may be omitted.
(Monitoring system)
FIG. 1 is a diagram showing an example of a schematic of a monitoring system according to an embodiment of the invention. A monitoring system 1 according to an embodiment of the present invention senses environmental information with one or a plurality of sensors provided on a target for abnormality determination. An example of an abnormality determination target is a power receiving and transforming facility such as a cubicle (high voltage power receiving facility).
The monitoring system 1 detects signs of failure by detecting movements different from normal times based on time-series data of sensor data obtained by sensing by each of one or a plurality of sensors. The monitoring system 1 detects a point of change (timing of change) in time-series data of sensor data when a sign of failure is detected.
The monitoring system 1 determines whether or not to issue an anomaly report based on sensor data obtained by sensing by each of the one or a plurality of sensors, based on the change point of the time-series data of the sensor data.

As shown in FIG. 1 , the monitoring system 1 includes a monitoring device 100 , a sensor server 200 , sensors 300 - 1 to 300 -n (n is an integer greater than 0), and a web server 400 . FIG. 1 shows a monitoring device 100, a sensor server 200, sensors 300-1 to 300-n, a web server 400, and a terminal device 500. FIG.
Monitoring device 100, sensor server 200, web server 400, and terminal device 500 communicate via network NW. The network NW includes, for example, the Internet, WAN (Wide Area Network), LAN (Local Area Network), provider equipment, wireless base stations, and the like.

Each of sensors 300 - 1 through 300 - n is connected to sensor server 200 . An example of each of sensors 300-1 through 300-n is an ultrasonic sensor, an ozone sensor, an odor sensor, or a sound wave sensor.
Ultrasonic sensors can detect changes with high sensitivity. For example, since an ultrasonic sensor can detect partial discharge on the surface of an insulator, high-voltage insulation deterioration can be detected. By being able to detect deterioration of high-voltage insulation, it is possible to check equipment for damage, fouling, corrosion, and abnormal noise, so it can replace sight and hearing.
The generated gas can be detected by an ozone sensor. For example, an ozone sensor can detect ozone generated in an organic insulator during discharge, thereby detecting high-voltage insulation deterioration. By being able to detect deterioration of high-voltage insulation, it is possible to check equipment for damage, fouling, corrosion, and abnormal noise, so it can replace sight and hearing.
An odor sensor can detect odor components. For example, an odor sensor can detect a bad odor when an insulator is overheated or burnt out, so a contact failure or overload can be detected. By being able to detect poor contact and overload, it is possible to check discoloration, corrosion, odors, and overheating of equipment, so it can replace sight, smell, and touch.
A sound wave sensor can detect changes in sounds that people perceive. For example, since the sound wave sensor can detect discharge sound and abnormal sound (vibration sound, beat), poor contact and overload can be detected. By detecting poor contact and overload, it is possible to check equipment for damage, contamination, corrosion, and abnormal noise, so it can replace sight and hearing.
By using an ultrasonic sensor, an ozone sensor, an odor sensor, and a sound wave sensor, five senses other than the human sense of taste can be substituted.

The sensor server 200 acquires detection values acquired by sensing by each of the sensors 300-1 to 300-n. An example of the detected value is analog data from 4 mA to 20 mA. The sensor server 200 acquires sensor data corresponding to each of the sensors 300-1 through 300-n by converting detection values acquired by each of the sensors 300-1 through 300-n into digital data. The sensor server 200 transmits the acquired sensor data corresponding to each of the sensors 300-1 to 300-n to the monitoring device 100. FIG.
The monitoring device 100 receives the sensor data corresponding to each of the sensors 300-1 to 300-n transmitted by the sensor server 200. FIG. Monitoring device 100 creates a sensor data notification including sensor data corresponding to each of sensors 300 - 1 to 300 -n received, and transmits the created sensor data notification to web server 400 .

Web server 400 receives the sensor data notification sent by monitoring device 100 . Based on one or more sensor data included in the received sensor data notification, the web server 400 receives an object such as an image for notifying the status of a monitoring target such as time-series data of each of one or more sensor data. to create
The terminal device 500 accesses the web server 400 to acquire an image for notifying the status of the monitoring target, and displays the acquired image for notifying the status of the monitoring target.

FIG. 2 is a diagram showing an example of the operation of the monitoring device included in the monitoring system according to this embodiment.
The monitoring device 100 creates time-series data of sensor data corresponding to each of the received sensors 300-1 to 300-n (1).
In monitoring device 100, change point detection AI acquires time-series data of sensor data corresponding to each of sensors 300-1 to 300-n. The change point detection AI determines whether or not there is a sensor that is moving differently than normal based on the time-series data of the sensor data corresponding to each of the acquired sensors 300-1 to 300-n. (2) to detect a sign of failure for the object equipped with the sensor. When the change point detection AI detects a sign of failure by determining that there is time-series data in which a movement different from normal times is present, it acquires the timing at which the movement different from normal times was detected.

In monitoring device 100, abnormality determination AI acquires time-series data of sensor data corresponding to each of sensors 300-1 to 300-n. The anomaly determination AI acquires information specifying the timing at which a motion different from normal is detected from the change point detection AI. The abnormality determination AI is based on the acquired time-series data of the sensor data corresponding to each of the sensors 300-1 to 300-n and the timing at which the movement different from the normal time is detected, and the sensor 300 for several hours before and after the timing. Based on the time-series data of the sensor data corresponding to each of the sensors 300-n to 300-n, it is determined whether or not an abnormality has occurred (3).
The abnormality determination AI determines whether or not to issue an abnormality report for notifying that an abnormality has occurred based on the determination result of whether or not an abnormality has occurred. When the monitoring device 100 determines to issue an anomaly report, the monitoring device 100 creates an anomaly occurrence notification and transmits the created anomaly occurrence notification to the web server 400 .

The web server 400 receives the abnormality occurrence notification sent by the monitoring device 100 . Web server 400 creates an object such as an image for notifying that an abnormality has occurred, based on the received abnormality occurrence notification.
The terminal device 500 acquires an image for notifying the state of the monitoring target from the web server 400, and displays the acquired image for notifying the state of the monitoring target. .
The sensor server 200 and the monitoring device 100 included in the monitoring system 1 will be described in detail below.

FIG. 3 is a diagram showing an example of a sensor server and a monitoring device included in the monitoring system according to this embodiment.
(Sensor server 200)
Sensor server 200 includes gateway 210 , sensor box 220 , and conversion box 230 .
Sensors 300-1 to 300-n are connected to conversion box 230. FIG. Conversion box 230 acquires detection values acquired by sensing by each of sensors 300-1 to 300-n.
The conversion box 230 acquires sensor data corresponding to each of the sensors 300-1 through 300-n by converting each detection value acquired by each of the sensors 300-1 through 300-n into digital data. . The conversion box 230 outputs to the sensor box 220 the acquired sensor data corresponding to each of the sensors 300-1 to 300-n.

The sensor box 220 acquires sensor data corresponding to each of the sensors 300-1 to 300-n output by the conversion box 230. FIG. The sensor box 220 outputs the acquired sensor data corresponding to each of the sensors 300-1 to 300-n to the gateway 210 according to a predetermined protocol. An example of the predetermined protocol is TCP/IP (Transmission Control Protocol/Internet Protocol). The sensor box 220 also includes an uninterruptible power supply (UPS).

The gateway 210 acquires sensor data corresponding to each of the sensors 300-1 to 300-n output by the sensor box 220. FIG. The gateway 210 transmits the acquired sensor data corresponding to each of the sensors 300-1 to 300-n to the monitoring device 100. FIG.
The gateway 210 communicates with external communication devices via the network NW. The gateway 210 communicates using a wireless communication method such as wireless LAN, Bluetooth (registered trademark), or LTE (registered trademark). The gateway 210 may communicate using a communication method such as a wired LAN, for example.
By providing an uninterruptible power supply in the sensor box 220, even if a power failure occurs and the supply of power to the sensor server 200 is stopped, power can be supplied from the uninterruptible power supply. Sensor data corresponding to each of 300-n can be sent to monitoring device 100. FIG.

(monitoring device 100)
The monitoring device 100 is implemented by a device such as a personal computer, server, smart phone, tablet computer, or industrial computer. An example of the monitoring device 100 is installed at a location away from power receiving and transforming equipment such as a cubicle (high voltage power receiving equipment). The monitoring device 100 includes, for example, a communication unit 110, a creation unit 120, a first processing unit 130-1, a second processing unit 130-2, an output unit 140, and a storage unit 150.

Communication unit 110 is realized by a communication module. The communication unit 110 communicates with an external communication device via the network NW. The communication unit 110 communicates using a wireless communication method such as a wireless LAN, Bluetooth (registered trademark), or LTE (registered trademark). The communication unit 110 may communicate using a communication method such as a wired LAN, for example. The communication unit 110 holds communication information necessary for communicating with the sensor server 200 and the web server 400 via the network NW.
Specifically, the communication unit 110 receives the sensor data transmitted by the sensor server 200 and corresponding to each of the sensors 300-1 to 300-n. The communication unit 110 acquires the sensor data notification output by the creation unit 120 and transmits the acquired sensor data notification to the web server 400 . The communication unit 110 acquires the abnormality occurrence notification output by the creation unit 120 and transmits the acquired abnormality occurrence notification to the web server 400 .
Creation unit 120 acquires sensor data corresponding to each of sensors 300-1 to 300-n received by communication unit 110. FIG. Creation unit 120 creates a sensor data notification destined for web server 400, including the acquired sensor data corresponding to each of sensors 300-1 to 300-n. Creation unit 120 outputs the created sensor data notification to communication unit 110 .

The first processing unit 130-1 includes a sensor data database (DB) 132-1. The first processing unit 130-1 acquires the sensor data corresponding to each of the sensors 300-1 to 300-n received by the communication unit 110. FIG. The first processing unit 130-1 stores the acquired sensor data corresponding to each of the sensors 300-1 to 300-n in the sensor data DB 132-1.
The first processing unit 130-1 corresponds to each of the sensors 300-1 to 300-n based on the sensor data corresponding to each of the sensors 300-1 to 300-n stored in the sensor data DB 132-1. Create time-series data of sensor data that
The first processing unit 130-1 machine-learns an environment in which an abnormality determination target is provided based on the created time-series data of the sensor data corresponding to each of the sensors 300-1 to 300-n. Here, the term "environment" refers to the totality of external factors such as nature surrounding humans and living things, which have some influence on the object of abnormality determination. For example, first processing unit 130-1 performs unsupervised learning based on time series data of sensor data corresponding to each of sensor 300-1 to sensor 300-n to obtain sensor 300-1 to sensor 300-n. Patterns between sensor data are learned for sensor data corresponding to each of n.
Based on the learning result, the first processing unit 130-1 converts the sensor data corresponding to each of the newly acquired sensors 300-1 to 300-n to the already acquired sensors 300-1 to 300-n. - Determine whether or not the sensor data can be classified into time-series data corresponding to each of n.

First processing unit 130-1 causes sensor data DB 132-1 to store newly acquired sensor data corresponding to each of sensors 300-1 to 300-n.
The first processing unit 130-1 corresponds to the newly acquired sensor data corresponding to each of the sensors 300-1 to 300-n and to each of the already acquired sensors 300-1 to 300-n. When it is determined that there is unclassifiable time-series data of the sensor data, it is determined that the unclassifiable newly acquired sensor data is abnormal.
When determining that the newly acquired sensor data is abnormal, the first processing unit 130-1 creates information specifying the acquisition timing of the sensor data determined to be abnormal. The first processing unit 130-1 transmits information specifying the timing at which the created sensor data determined to be abnormal and information specifying the sensor corresponding to the sensor data to the second processing unit 130-2. Output.

The second processing unit 130-2 has a learning model 132-2. The learning model 132-2 is based on the sensor data corresponding to each of the sensors 300-1 to 300-n and the abnormalities occurring in the targets provided with the sensors 300-1 to 300-n. -1 through sensor 300-n and machine learning of the relationship between the sensor data corresponding to each of the sensors 300-1 through 300-n and the information specifying whether or not the object provided with the sensor 300-1 through sensor 300-n is abnormal can get.
Based on the sensor data corresponding to each of the sensors 300-1 to 300-n, a person determines whether or not an abnormality has occurred in the object provided with the sensors 300-1 to 300-n. -1 to sensor 300-n and the relationship between the information specifying whether or not the object provided with sensor 300-1 to sensor 300-n is abnormal by machine learning , the learning model 132-2 may be updated.
The second processing unit 130-2 acquires the sensor data corresponding to each of the sensors 300-1 to 300-n received by the communication unit 110. FIG. The second processing unit 130-2 acquires information specifying the timing at which the sensor data output by the first processing unit 130-1 and determined to be abnormal was acquired, and information specifying the sensor corresponding to the sensor data. do.
Second processing unit 130-2 stores sensor data DB 132-1 for each of sensors 300-1 to 300-n based on the information specifying the timing at which sensor data determined to be abnormal is acquired. From the sensor data corresponding to , the sensor data corresponding to each of the sensors 300-1 to 300-n obtained several hours before and after the timing are obtained.
Based on the acquired sensor data corresponding to each of the sensors 300-1 to 300-n and the learning model 132-2, the second processing unit 130-2 detects the sensors 300-1 to 300-n. It is determined whether or not the detected object is abnormal.

The second processing unit 130-2 determines that an abnormality notification for notifying that an abnormality has occurred is performed when it is determined that the target provided with the sensors 300-1 to 300-n is abnormal. , and when it is determined that the target provided with the sensors 300-1 to 300-n is not abnormal, it is determined not to issue an abnormality report for notifying that an abnormality has occurred.
When second processing unit 130-2 determines to issue an anomaly report, creating unit 120 creates an anomaly occurrence notification addressed to web server 400, including information specifying that an anomaly has occurred. Creation unit 120 outputs the created abnormality notification to communication unit 110 .
The output unit 140 acquires the determination result as to whether or not the target is abnormal from the second processing unit 130-2, and outputs the acquired determination result as to whether or not the target is abnormal. For example, the output unit 140 may output the determination result as to whether or not the object is abnormal by displaying it on a display unit (not shown), or outputting it by voice.

The storage unit 150 is implemented by a HDD (Hard Disk Drive), flash memory, RAM (Random Access Memory), ROM (Read Only Memory), or the like.
Creation unit 120, first processing unit 130-1, second processing unit 130-2, and output unit 140 are configured by a hardware processor such as a CPU (Central Processing Unit) stored in storage unit 150. It is realized by executing a computer program (software).
Some or all of these functional units are hardware (circuits) such as LSI (Large Scale Integration), ASIC (Application Specific Integrated Circuit), FPGA (Field-Programmable Gate Array), and GPU (Graphics Processing Unit). (including circuit), or by cooperation between software and hardware. The computer program may be stored in advance in a storage device such as an HDD or flash memory, or may be stored in a removable storage medium such as a DVD or CD-ROM, and the storage medium may be loaded into a drive device. may be installed with

(Operation of monitoring system)
FIG. 4 is a flow chart showing an example of the operation of the monitoring system according to this embodiment. Here, as an example, in monitoring device 100, first processing unit 130-1 determines whether an abnormality is to be determined based on time-series data of sensor data corresponding to each of sensors 300-1 to 300-n. Unsupervised learning of the provided environment and relationship between sensor data corresponding to each of the sensors 300-1 to 300-n and patterns of sensor data corresponding to each of the sensors 300-1 to 300-n will be described.
Further, in the monitoring device 100, the second processing unit 130-2 generates sensor data corresponding to each of the sensors 300-1 to 300-n, and the sensor data generated by the target provided with the sensors 300-1 to 300-n. sensor data corresponding to each of the sensors 300-1 to 300-n and information specifying whether or not the target provided with the sensors 300-1 to 300-n is abnormal based on the abnormality A case will be described in which a learning model 132-2 obtained by machine-learning the relationship between is provided.

(Step S1-1)
Sensor 300-1 obtains a detection value by sensing.
(Step S2-1)
Sensor 300 - 1 transmits the acquired detection value to sensor server 200 .
In the sensors 300-2 to 300-n-1, the same processes as steps S1-1 to S2-1 are performed.
(Step S3-1)
Sensor 300-n obtains a detection value by sensing.
(Step S4-1)
The sensor 300-n transmits the acquired detection value to the sensor server 200. FIG.

(Step S5-1)
In sensor server 200, conversion box 230 acquires detection values transmitted by sensors 300-1 to 300-n.
(Step S6-1)
In the sensor server 200, the conversion box 230 A/D-converts the acquired detection value transmitted by each of the sensors 300-1 to 300-n.
(Step S7-1)
In the sensor server 200, the sensor box 220 acquires sensor data corresponding to each of the sensors 300-1 to 300-n obtained by conversion into digital data by the conversion box 230. FIG. The sensor box 220 outputs the acquired sensor data corresponding to each of the sensors 300-1 to 300-n to the gateway 210 according to a predetermined protocol.
(Step S8-1)
In the sensor server 200, the gateway 210 acquires the sensor data corresponding to each of the sensors 300-1 to 300-n output by the sensor box 220. FIG. The gateway 210 transmits the acquired sensor data corresponding to each of the sensors 300-1 to 300-n to the monitoring device 100. FIG.

(Step S9-1)
In the monitoring device 100, the communication unit 110 receives sensor data corresponding to each of the sensors 300-1 to 300-n transmitted by the sensor server 200. FIG.
(Step S10-1)
In the monitoring device 100, the creation unit 120 acquires the sensor data corresponding to each of the sensors 300-1 to 300-n, which are transmitted by the sensor server 200 and received by the communication unit 110. FIG. Creation unit 120 creates a sensor notification addressed to web server 400, including the acquired sensor data corresponding to each of sensors 300-1 to 300-n.
(Step S11-1)
In the monitoring device 100 , the creation unit 120 outputs the created sensor notification to the communication unit 110 . The communication unit 110 acquires the sensor notification output by the creation unit 120 and transmits the acquired sensor notification to the web server 400 .
Web server 400 receives the sensor data notification sent by monitoring device 100 . Based on one or more sensor data included in the received sensor data notification, the web server 400 receives an object such as an image for notifying the status of a monitoring target such as time-series data of each of one or more sensor data. to create
The terminal device 500 accesses the web server 400 to acquire an image for notifying the status of the monitoring target, and displays the acquired image for notifying the status of the monitoring target.

(Step S12-1)
In the monitoring device 100, the first processing unit 130-1 acquires sensor data corresponding to each of the sensors 300-1 to 300-n received by the communication unit 110. FIG. The first processing unit 130-1 stores the acquired sensor data corresponding to each of the sensors 300-1 to 300-n in the sensor data DB 132-1. The first processing unit 130-1 corresponds to each of the sensors 300-1 to 300-n based on the sensor data corresponding to each of the sensors 300-1 to 300-n stored in the sensor data DB 132-1. Create time-series data of sensor data that
(Step S13-1)
In the monitoring device 100, the first processing unit 130-1, based on the learning result, stores sensor data corresponding to each of the newly acquired sensors 300-1 to 300-n from the already acquired sensor 300. -1 to sensor 300-n, newly acquired sensor data corresponding to each of sensors 300-1 to 300-n is determined by determining whether it can be classified into time-series data of sensor data corresponding to each of sensors 300-1 to 300-n. determines whether or not there is an abnormality in If it is determined that all the newly acquired sensor data corresponding to each of the sensors 300-1 to 300-n are not abnormal, the process returns to step S9-1.
The first processing unit 130-1 corresponds to the newly acquired sensor data corresponding to each of the sensors 300-1 to 300-n and to each of the already acquired sensors 300-1 to 300-n. When it is determined that there is unclassifiable time-series data of the sensor data, it is determined that the unclassifiable newly acquired sensor data is abnormal.
When determining that the newly acquired sensor data is abnormal, the first processing unit 130-1 creates information specifying the acquisition timing of the sensor data determined to be abnormal. The first processing unit 130-1 transmits information specifying the timing at which the created sensor data determined to be abnormal and information specifying the sensor corresponding to the sensor data to the second processing unit 130-2. Output.

(Step S14-1)
In the monitoring device 100, the second processing unit 130-2 acquires the sensor data corresponding to each of the sensors 300-1 to 300-n received by the communication unit 110. FIG. When the first processing unit 130-1 determines that there is an abnormality, the second processing unit 130-2 acquires information specifying the timing at which the sensor data determined to be abnormal and the sensor data corresponding to the sensor data. from the first processing unit 130-1.
Second processing unit 130-2 stores sensor data DB 132-1 for each of sensors 300-1 to 300-n based on the information specifying the timing at which sensor data determined to be abnormal is obtained. From the sensor data corresponding to , the sensor data corresponding to each of the sensors 300-1 to 300-n acquired several hours before and after the timing are acquired.
Based on the acquired sensor data corresponding to each of the sensors 300-1 to 300-n and the learning model 132-2, the second processing unit 130-2 detects the sensors 300-1 to 300-n. It is determined whether or not the detected target is abnormal.
The second processing unit 130-2 determines that an abnormality notification for notifying that an abnormality has occurred is performed when it is determined that the target provided with the sensors 300-1 to 300-n is abnormal. , and when it is determined that the target provided with the sensors 300-1 to 300-n is not abnormal, it is determined not to issue an abnormality report for notifying that an abnormality has occurred. If it is determined that the target is not abnormal, the process returns to step S9-1.
(Step S15-1)
In the monitoring device 100, when the second processing unit 130-2 determines to issue an anomaly report, the creating unit 120 sends an anomaly occurrence address to the web server 400, which includes information specifying that an anomaly has occurred. Create notifications.

(Step S16-1)
In the monitoring device 100 , the creation unit 120 outputs the created abnormality notification to the communication unit 110 . The communication unit 110 acquires the abnormality occurrence notification output by the creation unit 120 and transmits the acquired abnormality occurrence notification to the web server 400 .
The web server 400 receives the abnormality occurrence notification sent by the monitoring device 100 . Web server 400 creates an object such as an image for notifying that an abnormality has occurred, based on the received abnormality occurrence notification.
The terminal device 500 acquires an image for notifying the state of the monitoring target from the web server 400, and displays the acquired image for notifying the state of the monitoring target. .

In the above-described embodiment, the first processing unit 130-1 stores the sensor data corresponding to each of the sensors 300-1 to 300-n acquired from the communication unit 110 in the sensor data DB 132-1, and causes the sensor data DB 132-1 to store the sensor data. Based on the sensor data corresponding to each of the sensors 300-1 to 300-n, time-series data of the sensor data corresponding to each of the sensors 300-1 to 300-n is created, and the created sensor 300- 1 to sensor 300-n, based on time-series data of sensor data corresponding to each of the sensors 300-n, a case has been described in which machine learning is performed on an environment in which an abnormality determination target is provided, but the present invention is not limited to this example.
For example, the first processing unit 130-1 may delete old sensor data among the sensor data corresponding to each of the sensors 300-1 to 300-n stored in the sensor data DB 132-1. That is, after learning sensor data corresponding to each of sensors 300-1 to 300-n, first processing unit 130-1 may preferentially use the most recent data for re-learning.
In the above-described embodiment, when it is determined that newly acquired sensor data is abnormal, the first processing unit 130-1 creates information specifying the acquisition timing of the sensor data determined to be abnormal. , the case where the information specifying the timing at which the created sensor data determined to be abnormal and the information specifying the sensor corresponding to the sensor data are output to the second processing unit 130-2 has been described. It is not limited to this example. For example, each of the sensors 300-1 to 300-n may output information regarding the detection time together with the detected value. In this case, when determining that the newly acquired sensor data is abnormal, first processing unit 130-1 acquires information specifying the detection time of the sensor data determined to be abnormal, and Information specifying the detection time of the sensor data determined to be abnormal may be used as information specifying the acquisition timing of the sensor data determined to be abnormal.

In the above-described embodiment, the second processing unit 130-2 selects the sensors 300-1 to 300 based on the sensor data corresponding to each of the sensors 300-1 to 300-n and the learning model 132-2. Although the case in which it is determined whether or not an object with -n is abnormal has been described, the present invention is not limited to this example.
For example, the second processing unit 130-2 selects the sensor data corresponding to each of the sensors 300-1 to 300-n, other than the sensor corresponding to the sensor data determined to be abnormal. , and learning model 132-2, it may be determined whether or not the target provided with sensor 300-1 to sensor 300-n is abnormal.
Further, for example, the second processing unit 130-2, based on the sensor data corresponding to at least a part of each of the sensors 300-1 to 300-n and the learning model 132-2, the sensor 300-1 to the sensor It may be determined whether the object provided with 300-n is abnormal.
In this case, the learning model 132-2 is based on sensor data corresponding to at least a part of each of the sensors 300-1 to 300-n, and an abnormality occurring in the object provided with the sensors 300-1 to 300-n. Identify whether the sensor data corresponding to at least a part of each of the sensors 300-1 to 300-n and the target provided with the sensors 300-1 to 300-n are abnormal based on It may be obtained by machine learning the relationship with the information to be used.
In the above-described embodiment, as an example, the case of monitoring small- and medium-sized electric facilities for private use has been described, but the present invention is not limited to this example. For example, it can be used for inspection systems in plants and factories, and monitoring systems for building security companies. By configuring in this way, it is possible to determine an abnormality and issue an alarm while suppressing the load of inspecting a plant or a factory by a person and the monitoring load of a building.

In the above-described embodiment, the monitoring system 1 may include one or a plurality of second sensors in addition to the sensors 300-1 to 300-n provided for the objects to be determined for abnormality. The first processing unit 130-1 may determine whether sensor data detected by each of the one or more second sensors is abnormal.
Second processing unit 130-2 further based on sensor data corresponding to at least part of each of the one or more second sensors and learning model 132-2, sensor 300-1 to sensor 300-n It may be determined whether the provided object is abnormal.
In this case, the learning model 132-2 includes sensor data corresponding to at least part of each of the sensors 300-1 to 300-n and sensor data corresponding to at least part of each of the one or more second sensors. , and an abnormality occurring in the object on which the sensors 300-1 to 300-n are provided, sensor data corresponding to at least a part of each of the sensors 300-1 to 300-n and one or more second sensors. Obtained by machine learning the relationship between sensor data corresponding to at least a part of each of the two sensors and information specifying whether or not the object provided with the sensors 300-1 to 300-n is abnormal. It may have been
In the embodiment described above, the case where one sensor server 200 is connected to the monitoring device 100 in the monitoring system 1 has been described, but the present invention is not limited to this example. For example, multiple sensor servers may be connected to the monitoring device 100 . A plurality of sensors may be connected to each of the plurality of sensor servers. A plurality of sensors may be provided for an abnormality determination target different from the sensors 300-1 to 300-n. In this case, the sensor data DB 132-1 and the learning model 132-2 may be provided for each target for abnormality determination. By configuring in this way, it is possible to determine whether or not to issue an anomaly report for each of the plurality of targets.

According to the monitoring system 1 according to the present embodiment, the monitoring system 1 includes one or more sensors 300-1 to 300-n provided for an object to determine abnormality, and one or more sensors 300-1 to 300-n. a first processing unit 130-1 for determining whether or not sensor data detected by each of n is abnormal; -1 to 300-n, and a second processing unit 130-2 that determines whether or not to report an abnormality based on the sensor data detected by each of 300-n.
By configuring in this way, the monitoring system 1 determines whether or not the sensor data detected by each of the one or more sensors 300-1 to 300-n provided for the abnormality determination target is abnormal. In addition, when it is determined that there is an abnormality, it is possible to determine whether or not to issue an abnormality report based on the sensor data detected by each of the one or more sensors 300-1 to 300-n.
For this reason, in addition to detecting changes different from normal times, it is determined whether or not to issue an anomaly report based on the sensor data detected by each of the one or more sensors 300-1 to 300-n, which is always acquired. I can judge.

In addition, since it is possible to improve the efficiency of on-site work and incorporate the knowledge of skilled engineers, it is possible to standardize the detection of signs of failure by AI. Combining sensors and AI can replace the five senses of engineers. Attach a sensor to the power receiving and transforming equipment of a private electric facility, detect changes in measured values, and let AI learn along with the failure determination results. In addition, it is possible to collate the results of inspections and measurements performed by engineers with the judgment results of AI to improve the evaluation and accuracy of AI.
Even in situations where engineers cannot go to the site due to disasters or infectious diseases, the status of electrical equipment can be checked remotely. In other words, it is possible to remotely detect signs of equipment deterioration in a highly accurate and uniform manner.
Since constantly changing measurement data is analyzed in chronological order, signs of equipment deterioration can be automatically detected without an on-site engineer.
Signs of deterioration can be uniformly detected with high accuracy regardless of the skill of the technician. In other words, highly efficient inspection and maintenance activities are possible.
Unnecessary power outages can be avoided by systematically dealing with equipment showing signs of failure.

In the monitoring system 1 described above, the first processing unit 130-1 preferentially uses the most recent sensor data after learning the sensor data detected by each of the one or more sensors 300-1 to 300-n. and relearn.
By configuring in this way, the monitoring system 1 preferentially uses the most recent sensor data after performing machine learning on the sensor data detected by each of the one or more sensors 300-1 to 300-n. Since re-learning is possible, it is judged whether or not the sensor data detected by each of the one or more sensors 300-1 to 300-n is abnormal compared to the case where the most recent sensor data is not preferentially used. Since the number (amount) of data used for determination can be reduced, the time required for determination can be shortened.
In the monitoring system 1 described above, the second processing unit 130-2 determines whether or not to issue an abnormality report based on sensor data detected by sensors other than the sensor that detected the sensor data determined to be abnormal. do.
By configuring in this way, the monitoring system 1 determines that any one of the one or more sensors 300-1 to 300-n provided for the abnormality determination target is abnormal. Based on the sensor data detected by sensors other than the sensor that detected the sensor data determined to be, it is possible to determine whether or not to issue an anomaly alarm. For this reason, in addition to detecting a change that differs from normal times, one or more of the sensors 300-1 to 300-n that are always acquired, other than the sensor that has detected the sensor data determined to be abnormal. Based on the sensor data detected by the sensor, it can be determined whether or not to report an abnormality.

In the above-described monitoring system 1, the first processing unit 130-1, based on the sensor data detected by each of the one or more sensors 300-1 to 300-n, each of one or more sensor data, It learns the relationship with the environment in which the target for abnormality determination is provided.
With this configuration, in the monitoring system 1, the first processing unit 130-1 detects one or more sensors based on sensor data detected by each of the one or more sensors 300-1 to 300-n. It is possible to learn the relationship between each of the sensor data and the environment in which the target for abnormality determination is provided. Specifically, first processing unit 130-1 performs unsupervised learning based on time-series data of sensor data corresponding to each of sensors 300-1 to 300-n. Therefore, first processing unit 130-1 can determine whether sensor data detected by each of one or more sensors 300-1 to 300-n is abnormal.

In the above-described monitoring system 1, the second processing unit 130-2, based on the sensor data detected by each of the one or more sensors 300-1 to 300-n, each of one or more sensor data, A relationship with each abnormality of a plurality of devices included in the target of abnormality determination is learned.
With this configuration, second processing unit 130-2 generates one or more sensors based on the sensor data detected by each of sensors 300-1 to 300-n and the abnormality that has occurred. The relationship between the sensor data detected by each of the sensors 300-1 to 300-n and the abnormality that has occurred can be machine-learned. Therefore, based on the sensor data detected by each of the one or more sensors 300-1 to 300-n, it is possible to determine whether or not to issue an anomaly report.

In the monitoring system 1 described above, the second processing unit 130-2 combines one or more sensor data based on the combination of sensor data detected by each of the one or more sensors 300-1 to 300-n. , and the relationship between each abnormality of a plurality of devices included in the abnormality determination target.
With this configuration, the second processing unit 130-2 can generate a single image based on a combination of sensor data detected by each of the one or more sensors 300-1 to 300-n and the abnormality that has occurred. Alternatively, the relationship between a combination of sensor data detected by each of the plurality of sensors 300-1 to 300-n and an abnormality that has occurred can be machine-learned. Therefore, out of the sensor data detected by each of the one or more sensors 300-1 to 300-n, based on the sensor data other than the sensor data detected by the sensor determined to be abnormal, an abnormality alarm is issued. You can decide whether to

In the monitoring system 1 described above, each of the one or more sensors is either an ultrasonic sensor, an ozone sensor, an odor sensor, or a sound wave sensor.
With this configuration, the monitoring system 1 includes one or more of the sensors 300-1 to 300-n, which are each of the sensors 300-1 to 300-n provided in the object to be judged for abnormality, an ultrasonic sensor, an ozone sensor, and an odor sensor. In addition to determining whether the sensor data detected by any of the sound wave sensors is abnormal, when it is determined that it is abnormal, each of the one or more sensors 300-1 to 300-n Based on the sensor data detected by , it is possible to determine whether or not to issue an anomaly alarm.
For this reason, in addition to detecting changes different from normal times, it is determined whether or not to issue an anomaly report based on sensor data detected by one or more sensors 300-1 to 300-n, which are always acquired. can.

In the above-described monitoring system 1, the monitoring system 1 further includes one or more second sensors other than the one or more sensors 300-1 to 300-n provided for the abnormality determination target. First processing unit 130-1 determines whether sensor data detected by each of the one or more second sensors is abnormal.
By configuring in this way, the monitoring system 1 can detect sensor data detected by each of the one or more second sensors other than the one or more sensors 300-1 to 300-n provided for the abnormality determination target. is abnormal or not. Therefore, it is possible to determine whether or not an environment other than an abnormality determination target environment is abnormal in addition to the abnormality determination target environment.

(Modification 1 of Embodiment)
A monitoring system 1a according to Modification 1 of the embodiment detects signs of failure by detecting movements different from normal times based on time-series data of sensor data obtained by sensing by each of one or more sensors. To detect. The monitoring system 1a detects a point of change (timing of change) in time series data of sensor data when a sign of failure is detected.
The monitoring system 1a determines a failure level based on sensor data obtained by sensing by each of one or more sensors, based on change points of time-series data of the sensor data. The monitoring system 1a determines whether or not to report an abnormality based on the failure level determination result.
FIG. 1 can be applied to the monitoring system 1a according to Modification 1 of the embodiment. However, it differs in that a monitoring device 100 a is provided instead of the monitoring device 100 .

The monitoring device 100a receives the sensor data corresponding to each of the sensors 300-1 to 300-n transmitted by the sensor server 200. FIG. Monitoring device 100 a creates a sensor data notification including sensor data corresponding to each of sensors 300 - 1 to 300 - n received, and transmits the created sensor data notification to web server 400 .
FIG. 5 is a diagram illustrating an example of the operation of a monitoring device included in the monitoring system according to Modification 1 of the embodiment;
The monitoring device 100a creates time-series data of sensor data corresponding to each of the received sensors 300-1 to 300-n (1).
In monitoring device 100a, change point detection AI acquires time series data of sensor data corresponding to each of sensors 300-1 to 300-n. The change point detection AI detects signs of failure by determining whether movement different from normal occurs based on time-series data of sensor data corresponding to each of the acquired sensors 300-1 to 300-n. Detect (2). When the change point detection AI detects a sign of failure by determining that there is time-series data in which a movement different from normal times is present, it acquires the timing at which the movement different from normal times was detected.

In the monitoring device 100a, the failure level determination AI acquires time-series data of sensor data corresponding to each of the sensors 300-1 to 300-n. The failure level determination AI acquires information specifying the timing at which a movement different from normal is detected from the change point detection AI.
Based on the time-series data of the sensor data corresponding to each of the acquired sensors 300-1 to 300-n and the timing at which movement different from normal time is detected, the failure level determination AI detects the sensor for several hours before and after that timing. Based on the time-series data of the sensor data corresponding to each of the sensors 300-1 to 300-n, the failure level is determined (3).
The abnormality determination AI determines whether or not to issue an abnormality report for notifying that an abnormality has occurred based on the determination result of the failure level. When the monitoring device 100 a determines to issue an anomaly report, the monitoring device 100 a creates an anomaly occurrence notification and transmits the created anomaly occurrence notification to the web server 400 .
The monitoring device 100a included in the monitoring system 1a will be described in detail below.

FIG. 6 is a diagram illustrating an example of a monitoring device included in a monitoring system according to Modification 1 of the embodiment;
(monitoring device 100a)
The monitoring device 100a is implemented by a device such as a personal computer, server, smart phone, tablet computer, or industrial computer. An example of the monitoring device 100a is installed at a location away from power receiving and transforming equipment such as a cubicle (high voltage power receiving equipment). The monitoring device 100a includes a communication unit 110, a creation unit 120a, a first processing unit 130-1, a second processing unit 130a-2, an output unit 140a, and a storage unit 150, for example.
The creating unit 120a acquires sensor data corresponding to each of the sensors 300-1 to 300-n received by the communication unit 110. FIG. Creation unit 120a creates a sensor data notification destined for web server 400, including the acquired sensor data corresponding to each of sensors 300-1 to 300-n. Creation unit 120 a outputs the created sensor data notification to communication unit 110 .

The second processing unit 130a-2 has a learning model 132a-2. The learning model 132a-2 is based on the sensor data corresponding to each of the sensors 300-1 to 300-n and the failure level of the object provided with the sensors 300-1 to 300-n. 1 to sensor 300-n, and the relationship between information specifying the failure level of the object provided with sensor 300-1 to sensor 300-n is obtained by machine learning. be done.
An example of machine learning is supervised learning. For example, a decision tree approach may be used for machine learning. A decision tree is a model that uses a tree structure to perform classification and regression. Values such as numerical values and classes are output from the root node, which is the origin, through conditional branching to the leaf node at the tip. Examples of failure levels are no failure, minor failure, medium failure, and major failure.
An example of a minor failure level is a failure that causes creeping discharge or corona discharge but does not leave a noticeable trace, in other words, a failure that cannot be visually detected. Occurrence of corona discharge is detected by changes in sensor detection values such as ultrasonic waves.
An example of a medium failure level is a failure at a level where creeping discharge or corona discharge occurs and traces can be confirmed on the insulator, in other words, there is a possibility of visual detection.
An example of a serious failure level is a level of failure in which creeping discharge or corona discharge occurs and clear traces can be confirmed on insulators and conductors (including grounding), in other words, failure at a level that can be visually detected.
Based on the sensor data corresponding to each of the sensors 300-1 to 300-n, a person determines the failure level of the target provided with the sensors 300-1 to 300-n. -n and the information specifying the failure level of the object provided with the sensors 300-1 to 300-n, the learning model 132a-2 is updated by machine learning. may

The second processing unit 130a-2 acquires the sensor data corresponding to each of the sensors 300-1 to 300-n received by the communication unit 110. FIG. The second processing unit 130a-2 acquires information specifying the timing at which the sensor data output by the first processing unit 130-1 and determined to be abnormal was acquired, and information specifying the sensor corresponding to the sensor data. do.
The second processing unit 130a-2 stores each of the sensors 300-1 to 300-n stored in the sensor data DB 132-1 based on the information specifying the timing at which the acquired sensor data determined to be abnormal. From the sensor data corresponding to , the sensor data corresponding to each of the sensors 300-1 to 300-n acquired several hours before and after the timing are acquired.
The second processing unit 130a-2, based on the acquired sensor data corresponding to each of the sensors 300-1 to 300-n and the learning model 132a-2, the sensor 300-1 to the sensor 300-n. determine the failure level of the target.

When the second processing unit 130a-2 determines that the failure level of the target provided with the sensor 300-1 to sensor 300-n is any one of a minor failure, a medium failure, and a major failure, an abnormality has occurred. When it is determined that the failure level of the target provided with the sensors 300-1 to 300-n is no failure, the occurrence of the abnormality is notified. It is determined that an anomaly notification will not be issued.
When the second processing unit 130a-2 determines to issue an abnormality report, the creation unit 120a includes information specifying that an abnormality has occurred and information specifying a target failure level. Create an anomaly notification addressed to . Creation unit 120 a outputs the created abnormality occurrence notification to communication unit 110 .
The output unit 140a acquires the target failure level determination result from the second processing unit 130a-2, and outputs the acquired target failure level determination result. For example, the output unit 140a may output the determination result of the target failure level by displaying it on a display unit (not shown) or outputting it by voice.

The creation unit 120a, the second processing unit 130a-2, and the output unit 140a are implemented by a hardware processor such as a CPU executing a computer program (software) stored in the storage unit 150, for example.
Also, some or all of these functional units may be implemented by hardware (including circuit units) such as LSI, ASIC, FPGA, GPU, etc., or by cooperation between software and hardware. may be implemented. The computer program may be stored in advance in a storage device such as an HDD or flash memory, or may be stored in a removable storage medium such as a DVD or CD-ROM, and the storage medium may be loaded into a drive device. may be installed with

(Operation of monitoring system)
FIG. 7 is a flow chart showing an example of the operation of the monitoring system according to Modification 1 of the present embodiment. Here, as an example, in the monitoring device 100a, the first processing unit 130-1 determines whether an abnormality is to be determined based on time-series data of sensor data corresponding to each of the sensors 300-1 to 300-n. Unsupervised learning of the provided environment and relationship between sensor data corresponding to each of the sensors 300-1 to 300-n and patterns of sensor data corresponding to each of the sensors 300-1 to 300-n will be described.
Further, in the monitoring device 100a, the second processing unit 130a-2 is provided with sensor data corresponding to at least a part of each of the sensors 300-1 to 300-n and the sensors 300-1 to 300-n. sensor data corresponding to at least a portion of each of the sensors 300-1 to 300-n and the failure level of the target provided with the sensors 300-1 to 300-n based on the target failure level A case will be described in which a learning model 132a-2 obtained by machine-learning the relationship with specified information is provided.
Since steps S1-1 to S13-1 described with reference to FIG. 4 can be applied to steps S1-2 to S13-2, description thereof will be omitted here.

(Step S14-2)
In the monitoring device 100a, the second processing unit 130a-2 acquires sensor data corresponding to each of the sensors 300-1 to 300-n received by the communication unit 110. FIG. The second processing unit 130a-2 acquires, from the first processing unit 130-1, information specifying the timing at which sensor data determined to be abnormal and information specifying the sensor corresponding to the sensor data are acquired. .
The second processing unit 130a-2 stores each of the sensors 300-1 to 300-n stored in the sensor data DB 132-1 based on the information specifying the timing at which the acquired sensor data determined to be abnormal. From the sensor data corresponding to , the sensor data corresponding to each of the sensors 300-1 to 300-n acquired several hours before and after the timing are acquired.
The second processing unit 130a-2, based on the acquired sensor data corresponding to each of the sensors 300-1 to 300-n and the learning model 132a-2, the sensor 300-1 to the sensor 300-n. determine the failure level of the target.
The second processing unit 130a-2 determines whether or not to issue an anomaly report based on the determination result of the failure level of the target equipped with the sensors 300-1 to 300-n. If it is determined not to issue an anomaly report, the process returns to step S9-2.

(Step S16-2)
In the monitoring device 100a, when the second processing unit 130a-2 determines that an abnormality report is to be issued, the creation unit 120a generates information specifying that an error has occurred and information specifying a target failure level. An abnormality occurrence notification addressed to the web server 400 is created.
(Step S17-2)
In the monitoring device 100 a , the creating unit 120 a outputs the created abnormality notification to the communication unit 110 . The communication unit 110 acquires the abnormality occurrence notification output by the creation unit 120 a and transmits the acquired abnormality occurrence notification to the web server 400 .
The web server 400 receives the abnormality notification sent by the monitoring device 100a. Web server 400 creates an object such as an image for notifying that an abnormality has occurred, based on the received abnormality occurrence notification.
The terminal device 500 acquires an image for notifying the state of the monitoring target from the web server 400, and displays the acquired image for notifying the state of the monitoring target. .

In the first modification of the above-described embodiment, the second processing unit 130a-2 receives sensor data corresponding to at least a part of each of the sensors 300-1 to 300-n and sensor data corresponding to at least a part of each of the sensors 300-1 to 300-n and the sensor 300 A learning model 132a-2 obtained by machine-learning the relationship with information specifying no failure, minor failure, moderate failure, and major failure as the target failure level provided with sensors 300-n from -1. Although the case has been described, it is not limited to this example.
For example, one to three failure levels may be applied, or five or more failure levels may be applied. Also, the failure level may or may not include no failure.

In the first modification of the above-described embodiment, the second processing unit 130a-2 uses the sensor data corresponding to each of the sensors 300-1 to 300-n and the learning model 132a-2 to determine the sensor 300- 1 to sensor 300-n have been described, but the present invention is not limited to this example.
For example, the second processing unit 130a-2, based on the sensor data corresponding to at least part of the acquired sensor data corresponding to each of the sensors 300-1 to 300-n and the learning model 132a-2, A fault level may be determined for the objects on which sensors 300-1 through 300-n are provided.
Further, for example, the second processing unit 130a-2 selects sensor data other than the sensor data corresponding to at least a part of each of the sensors 300-1 to 300-n and determined to be abnormal. Based on the sensors and the learning model 132a-2, the fault level for the object with which sensors 300-1 through 300-n are equipped may be determined.
In the modified example 1 of the above-described embodiment, the monitoring device 100a may include one or more second sensors in addition to the sensors 300-1 to 300-n provided for the abnormality determination targets.
Second processing unit 130a-2 further based on sensor data corresponding to at least part of each of the one or more second sensors and learning model 132a-2, sensor 300-1 to sensor 300-n A failure level of the provisioned object may be determined.
In this case, the learning model 132a-2 includes sensor data corresponding to at least part of each of the sensors 300-1 to 300-n and sensor data corresponding to at least part of each of the one or more second sensors. , and an abnormality occurring in the object on which the sensors 300-1 to 300-n are provided, sensor data corresponding to at least a part of each of the sensors 300-1 to 300-n and one or a plurality of second It was obtained by machine learning the relationship between sensor data corresponding to at least a part of each of the two sensors and information specifying the failure level of the target provided with the sensors 300-1 to 300-n. There may be.

According to the monitoring system 1a according to Modification 1 of the embodiment, in the monitoring system 1, the second processing unit 130a-2 detects based on a combination of sensor data detected by each of one or a plurality of sensors. Then, the relationship between the combination of one or more sensor data and the failure level of each of the plurality of devices included in the abnormality determination target is learned.
With this configuration, the second processing unit 130a-2 combines sensor data detected by each of the one or more sensors 300-1 to 300-n and A combination of sensor data corresponding to each of the sensors 300-1 to 300-n and a failure level of the target provided with the sensor 300-1 to the sensor 300-n are determined based on the failure level of the provided target. It is possible to machine-learn the relationship with the specified information.
Therefore, based on the sensor data detected by each of the one or more sensors 300-1 to 300-n, it is possible to determine whether or not to issue an anomaly report.
In addition, since it is possible to improve the efficiency of on-site work and incorporate the knowledge of skilled engineers, it is possible to standardize the detection of signs of failure by AI. Combining sensors and AI can replace the five senses of engineers. Attach a sensor to the power receiving and transforming equipment of a private electric facility, detect changes in measured values, and make AI learn along with the judgment result of the failure level. In addition, it is possible to collate the results of inspections and measurements performed by engineers with the judgment results of AI to improve the evaluation and accuracy of AI.

(Modification 2 of Embodiment)
FIG. 8 is a schematic diagram illustrating an example of a monitoring system according to Modification 2 of the embodiment. A monitoring system 1b according to Modification 2 of the embodiment senses environmental information with a plurality of sensors provided for each of a plurality of objects for which abnormality is to be determined. An example of a target for abnormality determination is a power receiving and transforming facility such as a cubicle.
The sensor servers 200-1 to 200-k are provided, for example, one for each of a plurality of targets for abnormality determination. The monitoring system 1b acquires sensor data obtained by sensing by each of one or more sensors provided for each of one or more targets via sensor servers 200-1 to 200-k. . Then, the monitoring system 1b detects a sign of failure in the target provided with the sensor that outputs the sensor data by detecting a movement different from normal in the time-series data of each sensor data. The monitoring system 1b detects a change point of time-series data of sensor data when a sign of failure is detected.
The monitoring system 1b detects sensor data obtained by sensing each of a plurality of sensors provided for a target having a sensor that detects sensor data determined to be abnormal based on a change point of time-series data of the sensor data. Based on the data, it is determined whether or not to issue an anomaly report.

As shown in FIG. 8, the monitoring system 1b includes a monitoring device 100b, sensor servers 200-1 to 200-k (k is an integer greater than 1), and sensor servers 200-1 to 200-k. sensor 300-1 to sensor 300-n (n is an integer of n>0) connected to each of k, and a web server 400. FIG. FIG. 8 shows monitoring device 100b, sensor servers 200-1 to 200-k, and sensors 300-1 to 300-n connected to sensor servers 200-1 to 200-k, respectively. , a web server 400, as well as a terminal device 500 are shown.
Monitoring device 100b, sensor servers 200-1 to 200-k, web server 400, and terminal device 500 communicate via network NW.
Each of sensor server 200-1 to sensor server 200-k can apply the above-described sensor server 200, so description thereof will be omitted here.

The monitoring device 100b receives sensor data corresponding to each of the sensors 300-1 to 300-n transmitted from each of the sensor servers 200-1 to 200-k. Monitoring device 100 b creates a sensor data notification including sensor data corresponding to each of sensors 300 - 1 to 300 - n received, and transmits the created sensor data notification to web server 400 .

The web server 400 receives the sensor data notification sent by the monitoring device 100b. Based on the multiple sensor data included in the received sensor data notification, the web server 400 monitors each of the sensor servers 200-1 to 200-k, such as time-series data of each of the multiple sensor data. Create an object such as an image to notify the status of
The terminal device 500 accesses the web server 400 to acquire an image for notifying the status of the monitoring target, and displays the acquired image for notifying the status of the monitoring target.

FIG. 2 can be applied to an example of the operation of the monitoring device included in the monitoring system according to Modification 2 of the embodiment.
The monitoring device 100b creates time-series data of sensor data corresponding to each of the sensors 300-1 to 300-n received for each of the sensor servers 200-1 to 200-k (1).
In monitoring device 100b, change point detection AI acquires time-series data of sensor data corresponding to each of sensors 300-1 to 300-n for each of sensor servers 200-1 to 200-k. The change point detection AI is based on time series data of sensor data corresponding to each of the sensors 300-1 to 300-n acquired for each of the sensor servers 200-1 to 200-k. By determining whether or not there is a sensor in which motion has occurred, a sign of failure is detected for the object equipped with the sensor (2). When the change point detection AI detects a sign of failure by determining that there is time-series data in which a movement different from normal times is present, it acquires the timing at which the movement different from normal times was detected.

In monitoring device 100b, abnormality determination AI acquires time-series data of sensor data corresponding to each of sensors 300-1 to 300-n for each of sensor servers 200-1 to 200-k. The anomaly determination AI obtains from the change point detection AI information specifying the timing at which a motion different from normal times was detected and information specifying the sensor that caused the motion different from normal times.
The anomaly determination AI determines the acquired sensors 300-1 to 300-n for each of the sensor servers 200-1 to 200-k based on the acquired information specifying the sensor that is moving differently from normal times. Acquire time-series data of sensor data corresponding to each of the sensors 300-1 to 300-n provided for the object equipped with the sensor that has caused a movement different from normal from the time-series data of the sensor data corresponding to each. do.
The abnormality determination AI is based on the acquired time-series data of the sensor data corresponding to each of the sensors 300-1 to 300-n and the timing at which the movement different from the normal time is detected, and the sensor 300 for several hours before and after the timing. Based on the time-series data of the sensor data corresponding to each of the sensors 300-n to 300-n, it is determined whether or not an abnormality has occurred (3).
The abnormality determination AI determines whether or not to issue an abnormality report for notifying that an abnormality has occurred based on the determination result of whether or not an abnormality has occurred. When the monitoring device 100 b determines to issue an anomaly report, it creates an anomaly occurrence notification including information for identifying the target in which an anomaly has occurred, and transmits the created anomaly occurrence notification to the web server 400 .

The web server 400 receives the abnormality notification sent by the monitoring device 100b. Web server 400 creates an object such as an image for notifying that an abnormality has occurred, based on the received abnormality occurrence notification.
The terminal device 500 acquires an image for notifying the state of the monitoring target from the web server 400, and displays the acquired image for notifying the state of the monitoring target. .
The monitoring device 100b included in the monitoring system 1b will be described in detail below.

FIG. 9 is a diagram illustrating an example of a monitoring device included in a monitoring system according to Modification 2 of the embodiment;
(monitoring device 100b)
The monitoring device 100b is implemented by a device such as a personal computer, server, smart phone, tablet computer, or industrial computer. An example of the monitoring device 100b is installed at a location away from power receiving and transforming equipment such as a cubicle (high voltage power receiving equipment). The monitoring device 100b includes a communication unit 110, a creation unit 120b, a first processing unit 130b-1, a second processing unit 130b-2, an output unit 140b, and a storage unit 150, for example.

The communication unit 110 holds communication information necessary for communicating with the sensor servers 200-1 to 200-k and the web server 400 via the network NW.
Specifically, communication unit 110 receives sensor data corresponding to each of sensors 300-1 to 300-n transmitted from each of sensor servers 200-1 to 200-k.
Creation unit 120b acquires sensor data corresponding to each of sensors 300-1 to 300-n, which are transmitted by each of sensor server 200-1 to sensor server 200-k received by communication unit 110. FIG. The creating unit 120b includes the acquired sensor data corresponding to each of the sensors 300-1 to 300-n transmitted by each of the sensor servers 200-1 to 200-k, with the web server 400 as the destination. Create sensor data notifications. Creation unit 120 b outputs the created sensor data notification to communication unit 110 .

The first processing unit 130b-1 has a sensor data database 132b-1. The first processing unit 130b-1 acquires sensor data corresponding to each of the sensors 300-1 to 300-n received by the communication unit 110 from each of the sensor servers 200-1 to 200-k. The first processing unit 130b-1 stores the acquired sensor data corresponding to each of the sensors 300-1 to 300-n for each of the sensor servers 200-1 to 200-k in the sensor data DB 132b-1. Memorize.
Based on the sensor data corresponding to each of the sensors 300-1 to 300-n for each of the sensor servers 200-1 to 200-k stored in the sensor data DB 132b-1, the first processing unit 130b-1 Then, time series data of sensor data corresponding to each of the sensors 300-1 to 300-n are created for each of the sensor servers 200-1 to 200-k.
Based on time-series data of sensor data corresponding to each of the sensors 300-1 to 300-n created for each of the sensor servers 200-1 to 200-k, the first processing unit 130b-1 performs sensor Machine learning is performed on an environment in which an abnormality determination target is provided for each of the server 200-1 to the sensor server 200-k. For example, the first processing unit 130b-1, for each of the sensor servers 200-1 to 200-k, based on time-series data of sensor data corresponding to each of the sensors 300-1 to 300-n, By performing unsupervised learning, patterns of sensor data corresponding to each of the sensors 300-1 to 300-n are learned.
For example, based on time-series data of sensor data detected by sensor 300-1 under the control of sensor server 200-1, patterns of sensor data are learned. Similarly, for each of the other sensors under the control of sensor server 200-1, patterns between sensor data in time-series data are learned. The same applies to each sensor under control of other sensor servers.
Based on the learning result, the first processing unit 130b-1 selects a sensor corresponding to each of the newly acquired sensors 300-1 to 300-n for each of the sensor servers 200-1 to 200-k. It is determined whether the data can be classified into time-series data of sensor data corresponding to each of the sensors 300-1 to 300-n that have already been acquired.

The first processing unit 130b-1 stores newly acquired sensor data corresponding to each of the sensors 300-1 to 300-n in the sensor data DB 132b-1 for each of the sensor servers 200-1 to 200-k. be memorized.
The first processing unit 130b-1 adds newly acquired sensor data corresponding to each of the sensors 300-1 to 300-n to each of the sensor servers 200-1 to 200-k. When it is determined that some of the time-series data of the sensor data corresponding to each of the sensors 300-1 to 300-n cannot be classified, it is determined that the newly acquired sensor data that cannot be classified is abnormal.
When the first processing unit 130b-1 determines that newly acquired sensor data is abnormal, the first processing unit 130b-1 creates information specifying the acquisition timing of the sensor data determined to be abnormal. The first processing unit 130b-1 transmits information specifying the timing at which the created sensor data determined to be abnormal and information specifying the sensor corresponding to the sensor data to the second processing unit 130b-2. Output.

The second processing unit 130b-2 has a learning model 132b-2. The learning model 132b-2 includes sensor data corresponding to each of the sensors 300-1 to 300-n and sensor data corresponding to each of the sensors 300-1 to 300-n for each of the sensor servers 200-1 to 200-k. Whether the sensor data corresponding to each of the sensors 300-1 to 300-n and the objects provided with the sensors 300-1 to 300-n are abnormal based on the abnormality occurring in the provided object It is obtained by machine learning the relationship with the information specifying whether or not. Alternatively, the learning model 132b-2 combines the sensor data from different sensors and the sensor data detected by each of the one or more sensors 300-1 to 300-n provided for the target, and determines whether the target is abnormal. It is obtained by acquiring and machine-learning the relationship with the information specifying whether or not it is for each of the plurality of objects.
For each of the sensor servers 200-1 to 200-k, based on the sensor data corresponding to each of the sensors 300-1 to 300-n, the target equipped with the sensors 300-1 to 300-n A person determines whether or not an abnormality has occurred, and determines whether the sensor data corresponding to each of the sensors 300-1 to 300-n and the objects provided with the sensors 300-1 to 300-n are abnormal. The learning model 132b-2 may be updated by machine-learning the relationship with the information specifying whether or not.
The second processing unit 130b-2 acquires sensor data corresponding to each of the sensors 300-1 to 300-n received by the communication unit 110 from each of the sensor servers 200-1 to 200-k. The second processing unit 130b-2 acquires information specifying the timing at which the sensor data output by the first processing unit 130b-1 and determined to be abnormal was acquired, and information specifying the sensor corresponding to the sensor data. do.
The second processing unit 130b-2 corresponds to each of the sensors 300-1 to 300-n acquired for each of the sensor servers 200-1 to 200-k based on the information specifying the acquired sensors. From the time-series data of the sensor data, time-series data of the sensor data corresponding to each of the sensors 300-1 to 300-n provided in the object equipped with the sensor corresponding to the sensor data determined to be abnormal is acquired. do.
The second processing unit 130b-2, based on the information specifying the acquisition timing of the sensor data determined to be abnormal, selects the acquired sensor data corresponding to each of the sensors 300-1 to 300-n. The sensor data corresponding to each of the sensors 300-1 to 300-n are acquired several hours before and after the timing.
Based on the acquired sensor data corresponding to each of sensors 300-1 to 300-n and the learning model 132b-2, second processing unit 130b-2 stores sensors 300-1 to 300-n. It is determined whether or not the detected target is abnormal.

The second processing unit 130b-2 determines that when it is determined that the target provided with the sensors 300-1 to 300-n is abnormal, an abnormality report is to be issued to notify that an abnormality has occurred. , and when it is determined that the target provided with the sensors 300-1 to 300-n is not abnormal, it is determined not to issue an abnormality report for notifying that an abnormality has occurred.
When the second processing unit 130b-2 determines to issue an anomaly report, the creating unit 120b includes information specifying that an anomaly has occurred and information identifying a target in which the anomaly has occurred. Create an anomaly notification addressed to . Creation unit 120 b outputs the created abnormality occurrence notification to communication unit 110 .
The output unit 140b acquires the information identifying the target and the determination result as to whether or not the target is abnormal from the second processing unit 130b-2, and obtains the acquired information identifying the target and whether or not the target is abnormal. and output the determination result. For example, the output unit 140b may output the information identifying the target and the determination result as to whether or not the target is abnormal by displaying them on a display unit (not shown), or outputting them by voice. good.

The first processing unit 130b-1 and the second processing unit 130b-2 are realized by executing a computer program (software) stored in the storage unit 150 by a hardware processor such as a CPU.
Also, some or all of these functional units may be implemented by hardware (including circuit units) such as LSI, ASIC, FPGA, GPU, etc., or by cooperation between software and hardware. may be implemented. The computer program may be stored in advance in a storage device such as an HDD or flash memory, or may be stored in a removable storage medium such as a DVD or CD-ROM, and the storage medium may be loaded into a drive device. may be installed with

(Operation of monitoring system)
FIG. 10 is a flowchart illustrating an example of the operation of the monitoring system according to Modification 2 of the embodiment. Operation after each of the sensors 300-1 to 300-n provided in each of the sensor servers 200-1 to 200-k transmits a detected value from the sensor server 200-1 to each of the sensor servers 200-k will be explained.
Further, here, as an example, in the monitoring device 100b, the first processing unit 130b-1 receives data from each of the sensors 300-1 to 300-n transmitted by each of the sensor servers 200-1 to 200-k. Based on the time-series data of the sensor data corresponding to the unsupervised learning of the environment in which the target for determining abnormality is provided. Specifically, for each of the sensor servers 200-1 to 200-k, sensor data corresponding to each of the sensors 300-1 to 300-n and data corresponding to each of the sensors 300-1 to 300-n A case in which the relationship between patterns of corresponding sensor data is learned will be described.
Further, the second processing unit 130b-2 generates sensor data corresponding to each of the sensors 300-1 to 300-n for each of the sensor servers 200-1 to 200-k, and Sensor data corresponding to each of the sensors 300-1 to 300-n and objects equipped with the sensors 300-1 to 300-n are stored based on an abnormality occurring in the target equipped with the sensor 300-n. A case will be described in which the learning model 132b-2 is obtained by machine learning the relationship with the information specifying whether or not there is an abnormality.

(Step S1-3)
In sensor server 200-1, conversion box 230 acquires detection values transmitted by sensors 300-1 to 300-n.
(Step S2-3)
In sensor server 200-1, conversion box 230 A/D-converts the acquired detection values transmitted by sensors 300-1 to 300-n.
(Step S3-3)
In sensor server 200-1, sensor box 220 acquires sensor data corresponding to each of sensors 300-1 to 300-n obtained by conversion into digital data by conversion box 230. FIG. The sensor box 220 outputs the acquired sensor data corresponding to each of the sensors 300-1 to 300-n to the gateway 210 according to a predetermined protocol.
(Step S4-3)
In the sensor server 200-1, the gateway 210 acquires the sensor data corresponding to each of the sensors 300-1 to 300-n output by the sensor box 220. FIG. The gateway 210 transmits the acquired sensor data corresponding to each of the sensors 300-1 to 300-n to the monitoring device 100b.
In the sensor server 200-2 to the sensor server 200-k-1, the same processes as steps S1-3 to S4-3 are performed.

(Step S5-3)
In sensor server 200-k, conversion box 230 acquires the detection values transmitted by each of sensors 300-1 to 300-n.
(Step S6-3)
In sensor server 200-k, conversion box 230 A/D-converts the acquired detection values transmitted by sensors 300-1 to 300-n.
(Step S7-3)
In sensor server 200-k, sensor box 220 acquires sensor data corresponding to each of sensors 300-1 to 300-n obtained by conversion into digital data by conversion box 230. FIG. The sensor box 220 outputs the acquired sensor data corresponding to each of the sensors 300-1 to 300-n to the gateway 210 according to a predetermined protocol.
(Step S8-3)
In the sensor server 200-k, the gateway 210 acquires the sensor data corresponding to each of the sensors 300-1 to 300-n output by the sensor box 220. FIG. The gateway 210 transmits the acquired sensor data corresponding to each of the sensors 300-1 to 300-n to the monitoring device 100b.

(Step S9-3)
In monitoring device 100b, communication unit 110 receives sensor data corresponding to each of sensors 300-1 to 300-n transmitted from each of sensor servers 200-1 to 200-k.
(Step S10-3)
In the monitoring device 100b, the creation unit 120b acquires the sensor data corresponding to each of the sensors 300-1 to 300-n transmitted by each of the sensor servers 200-1 to 200-k received by the communication unit 110. do. Creation unit 120b generates a sensor notification addressed to web server 400, including sensor data corresponding to each of sensors 300-1 to 300-n, for each of acquired sensor servers 200-1 to 200-k. to create
(Step S11-3)
In the monitoring device 100 b , the creation unit 120 b outputs the created sensor notification to the communication unit 110 . The communication unit 110 acquires the sensor notification output by the creation unit 120 b and transmits the acquired sensor notification to the web server 400 .
The web server 400 receives the sensor data notification sent by the monitoring device 100b. Based on the sensor data corresponding to each of the sensors 300-1 to 300-n for each of the sensor servers 200-1 to 200-k included in the received sensor data notification, the web server 400 For each of the sensor servers 200-1 to 200-k, an object such as an image for notifying the state of the monitored object such as time-series data of each of a plurality of sensor data is created.
The terminal device 500 accesses the web server 400 to acquire an image for notifying the status of the monitoring target, and displays the acquired image for notifying the status of the monitoring target.

(Step S12-3)
In the monitoring device 100b, the first processing unit 130b-1 corresponds to each of the sensors 300-1 to 300-n for each of the sensor servers 200-1 to 200-k received by the communication unit 110. Get sensor data. The first processing unit 130b-1 stores the acquired sensor data corresponding to each of the sensors 300-1 to 300-n for each of the sensor servers 200-1 to 200-k in the sensor data DB 132b-1. be memorized. The first processing unit 130b-1 stores sensor data corresponding to each of the sensors 300-1 to 300-n for each of the sensor servers 200-1 to 200-k stored in the sensor data DB 132b-1. Based on this, time-series data of sensor data corresponding to each of the sensors 300-1 to 300-n is created.
(Step S13-3)
In the monitoring device 100b, the first processing unit 130b-1, based on the learning result, calculates the newly acquired sensor 300-1 to sensor 300-n for each of the sensor servers 200-1 to 200-k. By determining whether or not the sensor data corresponding to each can be classified into time-series data of sensor data corresponding to each of the sensors 300-1 to 300-n that have already been acquired, the sensor server 200-1 to the sensor server 200-k, it is determined whether or not the newly acquired sensor data corresponding to each of the sensors 300-1 to 300-n includes an abnormality. If it is determined that all of the newly acquired sensor data corresponding to each of the sensors 300-1 to 300-n are not abnormal, the process returns to step S9-3.
The first processing unit 130b-1 adds newly acquired sensor data corresponding to each of the sensors 300-1 to 300-n to each of the sensor servers 200-1 to 200-k. When it is determined that some of the time-series data of the sensor data corresponding to each of the sensors 300-1 to 300-n cannot be classified, it is determined that the newly acquired sensor data that cannot be classified is abnormal.
When the first processing unit 130b-1 determines that newly acquired sensor data is abnormal, the first processing unit 130b-1 creates information specifying the acquisition timing of the sensor data determined to be abnormal. The first processing unit 130b-1 transmits information specifying the timing at which the created sensor data determined to be abnormal and information specifying the sensor corresponding to the sensor data to the second processing unit 130b-2. Output.

(Step S14-3)
In the monitoring device 100b, the second processing unit 130b-2 corresponds to each of the sensors 300-1 to 300-n transmitted by each of the sensor servers 200-1 to 200-k received by the communication unit 110. Get sensor data. When the first processing unit 130b-1 determines that the first processing unit 130b-1 is abnormal, the second processing unit 130b-2 acquires information specifying the timing at which the sensor data determined to be abnormal and the sensor data corresponding to the sensor data. from the first processing unit 130b-1.
The second processing unit 130b-2 corresponds to each of the sensors 300-1 to 300-n acquired for each of the sensor servers 200-1 to 200-k based on the information specifying the acquired sensors. From the time-series data of the sensor data, time-series data of the sensor data corresponding to each of the sensors 300-1 to 300-n provided in the object equipped with the sensor corresponding to the sensor data determined to be abnormal is acquired. do.
The second processing unit 130b-2 extracts the acquired sensor data corresponding to each of the sensors 300-1 to 300-n based on the information specifying the acquisition timing of the acquired sensor data determined to be abnormal. , sensor data corresponding to each of the sensors 300-1 to 300-n are acquired several hours before and after the timing.
Second processing unit 130b-2 acquires sensor data corresponding to each of sensors 300-1 to 300-n and based on learning model 132b-2, sensor 300-1 to sensor 300-n are provided. It is determined whether or not the detected target is abnormal.
The second processing unit 130b-2 determines that when it is determined that the target provided with the sensors 300-1 to 300-n is abnormal, an abnormality report is to be issued to notify that an abnormality has occurred. , and when it is determined that the target provided with the sensors 300-1 to 300-n is not abnormal, it is determined not to issue an abnormality report for notifying that an abnormality has occurred. If it is determined that the target is not abnormal, the process returns to step S9-3.
(Step S15-3)
In the monitoring device 100b, when the second processing unit 130b-2 determines to issue an anomaly report, the creating unit 120b sends an anomaly occurrence address to the web server 400, which includes information specifying that an anomaly has occurred. Create notifications.

(Step S16-3)
In the monitoring device 100 b , the creating unit 120 b outputs the created abnormality notification to the communication unit 110 . The communication unit 110 acquires the abnormality occurrence notification output by the creation unit 120 b and transmits the acquired abnormality occurrence notification to the web server 400 .
The web server 400 receives the abnormality notification sent by the monitoring device 100b. Web server 400 creates an object such as an image for notifying that an abnormality has occurred, based on the received abnormality occurrence notification.
The terminal device 500 acquires an image for notifying the state of the monitoring target from the web server 400 and displays the acquired image for notifying the state of the monitoring target. .
In the flowchart shown in FIG. 10, the monitoring device 100b may perform the above-described processing based on sensor data transmitted by at least one of the sensor servers 200-1 to 200-k.

In the modification 2 of the above-described embodiment, as an example, the case where the sensor servers 200-1 to 200-k are provided for each of a plurality of targets for abnormality determination was described. Examples are not limited. For example, one sensor server may be provided for a plurality of targets for abnormality determination, or a plurality of sensor servers may be provided for a single plurality of targets for abnormality determination.
In the second modification of the above-described embodiment, the first processing unit 130b-1 acquires from the communication unit 110 the sensors 300-1 to 300-n for each of the sensor servers 200-1 to 200-k. are stored in the sensor data DB 132b-1, and corresponding to each of the sensors 300-1 to 300-n for each of the stored sensor servers 200-1 to 200-k. Based on the sensor data obtained, time series data of the sensor data corresponding to each of the sensors 300-1 to 300-n is created for each of the sensor servers 200-1 to 200-k, and the created sensor server 200-1 to sensor server 200-k, based on time-series data of sensor data corresponding to each of sensors 300-1 to 300-n, an environment provided with a target for abnormality determination is machined. Although the case of learning has been described, the present invention is not limited to this example.
For example, the first processing unit 130b-1 corresponds to each of the sensors 300-1 to 300-n for each of the sensor servers 200-1 to 200-k stored in the sensor data DB 132b-1. Of the sensor data, old sensor data may be erased. That is, the first processing unit 130b-1 learns the sensor data corresponding to each of the sensors 300-1 to 300-n for each of the sensor servers 200-1 to 200-k, and then, after learning the sensor data corresponding to each of the sensors 300-1 to 300-n, may be preferentially used and re-learned.

In the above-described embodiment, when the first processing unit 130b-1 determines that newly acquired sensor data is abnormal, the first processing unit 130b-1 creates information specifying the acquisition timing of the sensor data determined to be abnormal. , the case of outputting to the second processing unit 130b-2 the information specifying the timing at which the created sensor data determined to be abnormal and the information specifying the sensor corresponding to the sensor data are output to the second processing unit 130b-2. It is not limited to this example. For example, each of the sensors 300-1 to 300-n may output information regarding the detection time together with the detected value. In this case, when the first processing unit 130b-1 determines that the newly acquired sensor data is abnormal, the first processing unit 130b-1 acquires information specifying the detection time of the sensor data determined to be abnormal, and Information specifying the detection time of the sensor data determined to be abnormal may be used as information specifying the acquisition timing of the sensor data determined to be abnormal.
In the modified example 2 of the above-described embodiment, the second processing unit 130b-2 detects the sensor 300 provided in the object having the sensor corresponding to the sensor data determined to be abnormal by the first processing unit 130b-1. -1 to sensor 300-n and based on the learning model 132b-2, it is determined whether or not the object provided with the sensor 300-1 to sensor 300-n is abnormal. Although the case has been described, it is not limited to this example.
For example, the second processing unit 130b-2 selects the sensor 300-1 to sensor 300-n provided in the object having the sensor corresponding to the sensor data determined to be abnormal by the first processing unit 130b-1. Of the sensor data corresponding to each, sensor data corresponding to sensors other than the sensor corresponding to the sensor data determined to be abnormal, and based on the learning model 132b-2, sensor 300-1 to sensor 300-n may be determined whether or not the object provided with is abnormal.
In addition, for example, the second processing unit 130b-2 detects the sensor data from the sensor 300-1 to the sensor 300- provided in the object having the sensor corresponding to the sensor data determined to be abnormal by the first processing unit 130b-1. n, and the learning model 132b-2, it may be determined whether or not the target provided with the sensors 300-1 to 300-n is abnormal. .
In this case, the learning model 132b-2 stores sensor data corresponding to at least a part of each of the sensors 300-1 to 300-n and sensor 300-n for each of the sensor servers 200-1 to 200-k. 1 to sensor 300-n, sensor data corresponding to at least a part of each of the sensors 300-1 to 300-n, and sensor 300-1 to sensor 300- It may be obtained by machine learning of the relationship with information specifying whether the object provided with n is abnormal or not.

In Modified Example 2 of the above-described embodiment, the monitoring device 100b may perform the above-described processing based on sensor data transmitted by at least one of the sensor servers 200-1 to 200-k.
In the modified example 2 of the above-described embodiment, the monitoring system 1b includes one or more second sensors in addition to the sensors 300-1 to 300-n provided for each of the plurality of targets for abnormality determination. good too. The first processing unit 130b-1 may determine whether sensor data detected by each of the one or more second sensors is abnormal.
The second processing unit 130b-2 is further based on sensor data corresponding to at least a part of each of one or more second sensors provided for each of the plurality of targets for abnormality determination, and the learning model 132b-2. Then, it may be determined whether or not the target provided with the sensors 300-1 to 300-n is abnormal.
In this case, the learning model 132b-2, for each of the sensor servers 200-1 to 200-k, sensor data corresponding to at least part of each of the sensors 300-1 to 300-n and one or more Each of the sensors 300-1 to 300-n based on sensor data corresponding to at least a part of each of the second sensors and an abnormality occurring in the object provided with the sensors 300-1 to 300-n Sensor data corresponding to at least a part of and sensor data corresponding to at least a part of each of the one or more second sensors, and whether the target provided with the sensor 300-1 to sensor 300-n is abnormal It may be obtained by machine learning of the relationship with information specifying whether.
In Modified Example 2 of the above-described embodiment, the number of sensors provided in each of sensor server 200-1 to sensor server 200-k may be different.

According to the monitoring system 1b according to the modified example 2 of the embodiment, the monitoring system 1b includes one or more sensors 300-1 to 300-n provided for the object to determine abnormality, and one or more sensors provided for the object. A first processing unit 130b-1 for determining whether or not sensor data detected by each of the sensors 300-1 to 300-n is abnormal for each sensor; a second processing unit 130b-2 for determining whether or not to issue an anomaly report based on sensor data detected by each of the plurality of sensors 300-1 to 300-n when determined; The unit 130b-2 learns the relationship between the sensor data detected by the sensors 300-1 to 300-n provided for each of the plurality of targets and the abnormality related to the target, and uses the result of the learning. , determines whether or not to issue an abnormality report for a target having a sensor that outputs sensor data determined to be abnormal by the first processing unit 130b-1.
With this configuration, the monitoring system 1b determines whether or not the sensor data detected by each of the one or more sensors 300-1 to 300-n provided for the abnormality determination target is abnormal. In addition, when it is determined to be abnormal, each of one or a plurality of sensors 300-1 to 300-n provided in the object equipped with the sensor that detected the sensor data determined to be abnormal detected Based on the sensor data, it can be determined whether or not to issue an anomaly alarm.
For this reason, in addition to detecting changes different from normal times, one or a plurality of sensors 300- provided in the subject equipped with sensors that have detected sensor data determined to be abnormal by detecting changes different from normal times. 1 to sensor 300-n, it is possible to determine whether or not to issue an anomaly report, so that determination accuracy can be improved.
If the second processing unit 130b-2 learns the relationship between the sensor data detected by each of the sensors 300-1 to 300-n provided for a specific target and the abnormality related to the specific target, , the learning result (learning model) is highly influenced by the specific target environment. Therefore, the versatility of the learning model is reduced. In the monitoring system 1b according to the modified example 2 of the present embodiment, the second processing unit 130b-2 detects sensor data detected by the sensors 300-1 to 300-n provided for each of a plurality of targets and anomalies related to the targets. Because it learns the relationship with , it can be highly versatile. As a result, the learning result (learning model) is used to improve the accuracy of determining whether or not to issue an abnormality report for a target having a sensor that outputs sensor data determined to be abnormal by the first processing unit 130b-1. can be high

In the above-described monitoring system 1b, the first processing unit 130b-1 learns the sensor data detected by each of the plurality of sensors 300-1 to 300-n, and then preferentially uses the most recent sensor data for reuse. learn.
By configuring in this way, the monitoring system 1b performs machine learning on the sensor data detected by each of the plurality of sensors 300-1 to 300-n, and then re-learns by preferentially using the most recent sensor data. Therefore, it is used to determine whether or not the sensor data detected by each of the plurality of sensors 300-1 to 300-n is abnormal compared to the case where the most recent sensor data is not preferentially used. Since the number (amount) of data can be reduced, the time required for determination can be shortened.
In the monitoring system 1b described above, the second processing unit 130b-2 detects sensor data detected by each of one or a plurality of sensors provided for a target equipped with a sensor that has detected sensor data determined to be abnormal, Based on the sensor data detected by sensors other than the sensor that detected the sensor data determined to be abnormal, it is determined whether or not to issue an abnormality alarm.
By configuring in this way, the monitoring system 1b determines that any one of the one or more sensors 300-1 to 300-n provided for each of the one or more objects for which abnormality is to be determined is abnormal. In this case, among the plurality of sensors 300-1 to 300-n provided to the object equipped with the sensor that detected the sensor data determined to be abnormal, the sensor other than the sensor that detected the sensor data determined to be abnormal Based on the sensor data detected by the sensor, it is possible to determine whether or not to report an abnormality.
For this reason, in addition to detecting a change different from normal time, from the plurality of sensors 300-1 provided in the object equipped with the sensor that detected the sensor data determined to be abnormal by detecting the change different from normal time Based on the sensor data detected by the sensors other than the sensor that detected the sensor data determined to be abnormal among the sensors 300-n, it is possible to determine whether or not to issue an abnormality report, thereby improving the determination accuracy.

In the monitoring system 1b described above, the first processing unit 130b-1, based on the sensor data detected by each of the one or more sensors 300-1 to 300-n, each of one or more sensor data, It learns the relationship with the environment in which the target for abnormality determination is provided.
With this configuration, in the monitoring system 1b, the first processing unit 130b-1 is configured to provide one or more sensors 300-1 to 300-n for each of the sensor servers 200-1 to 200-k. can learn the relationship between each of the one or a plurality of sensor data and the environment in which the abnormality determination target is provided based on the sensor data detected by each of the. Specifically, the first processing unit 130b-1 converts sensor data corresponding to each of the sensors 300-1 to 300-n into time series data for each of the sensor servers 200-1 to 200-k. based on unsupervised learning. Therefore, the first processing unit 130b-1 can determine whether or not the sensor data detected by each of the plurality of sensors 300-1 to 300-n is abnormal.

In the monitoring system 1b described above, the second processing unit 130b-2 is included in the targets for determining an abnormality determined based on sensor data detected by one or more sensors provided in each of the plurality of targets. The relationship between each of the one or more sensor data and the abnormality of each of the plurality of devices included in the abnormality determination target is updated based on the state of each of the plurality of devices included in the device.
With this configuration, it is possible to update the relationship between each of the one or more sensor data and the abnormality of each of the plurality of devices included in the abnormality determination target, so that it is possible to determine whether or not to issue an abnormality alarm. Can improve accuracy.

In the above-described monitoring system 1b, the second processing unit 130b-2 is based on a combination of sensor data detected by each of one or a plurality of sensors 300-1 to 300-n provided for each of a plurality of targets. , a relationship between a combination of one or more sensor data and an abnormality in each of a plurality of devices included in each of a plurality of targets for abnormality determination.
With this configuration, the second processing unit 130b-2 generates one or The relationship between the combination of sensor data detected by each of the plurality of sensors 300-1 to 300-n and the abnormality that has occurred can be machine-learned. Therefore, based on the sensor data detected by each of the plurality of sensors 300-1 to 300-n, it is possible to determine whether or not to issue an abnormality report.

In the above-described monitoring system 1b, the monitoring system 1b further includes one or more second sensors other than the one or more sensors 300-1 to 300-n provided for the object of abnormality determination. The second processing unit 130b-2 determines whether sensor data detected by each of the one or more second sensors is abnormal.
By configuring in this way, the monitoring system 1b can detect sensor data detected by each of the one or more second sensors other than the one or more sensors 300-1 to 300-n provided for the abnormality determination target. is abnormal or not. Therefore, it is possible to determine whether or not an environment other than an abnormality determination target environment is abnormal in addition to the abnormality determination target environment.

(Modification 3 of Embodiment)
A monitoring system 1c according to Modification 3 of the embodiment obtains information by sensing each of one or more targets with respect to each of one or more targets via sensor servers 200-1 to 200-k. Acquire sensor data Then, the monitoring system 1c detects a sign of failure in the target provided with the sensor that outputs the sensor data by detecting movement different from normal time in the time-series data of each sensor data. The monitoring system 1c detects a point of change (timing of change) in time series data of sensor data when a sign of failure is detected.
The monitoring system 1c detects sensor data obtained by sensing each of a plurality of sensors provided for an object having a sensor that detects sensor data determined to be abnormal based on a change point of time-series data of the sensor data. A fault level is determined based on the data. The monitoring system 1c determines whether or not to issue an anomaly report based on the failure level determination result. Note that the failure level is assumed to be equivalent to the failure level described in the embodiment.
FIG. 8 can be applied to the monitoring system 1c according to Modification 3 of the embodiment. However, it differs in that a monitoring device 100c is provided instead of the monitoring device 100b.

The monitoring device 100c receives sensor data corresponding to each of the sensors 300-1 to 300-n transmitted from each of the sensor servers 200-1 to 200-k. The monitoring device 100c creates a sensor data notification including sensor data corresponding to each of the sensors 300-1 to 300-n received from each of the sensor servers 200-1 to 200-k, and sends the created sensor data A notification is sent to the web server 400 .
FIG. 5 can be applied to an example of the operation of the monitoring device included in the monitoring system 1c according to Modification 3 of the embodiment.
The monitoring device 100c creates time series data of sensor data corresponding to each of the sensors 300-1 to 300-n received from each of the sensor servers 200-1 to 200-k (1).
In monitoring device 100c, change point detection AI acquires time-series data of sensor data corresponding to each of sensors 300-1 to 300-n for each of sensor servers 200-1 to 200-k. The change point detection AI is based on the time-series data of the sensor data corresponding to each of the sensors 300-1 to 300-n acquired for each of the sensor servers 200-1 to 200-k. By determining whether or not there is a sensor in which movement has occurred, a sign of failure is detected for the object equipped with the sensor (2). When the change point detection AI detects a sign of failure by determining that there is time-series data in which a movement different from normal times is present, it acquires the timing at which the movement different from normal times was detected.

In the monitoring device 100c, the failure level determination AI acquires time-series data of sensor data corresponding to each of the sensors 300-1 to 300-n for each of the sensor servers 200-1 to 200-k. The failure level determination AI obtains from the change point detection AI information specifying the timing at which a motion different from normal times was detected and information specifying the sensor that caused the motion different from normal times.
The failure level determination AI determines the acquired sensor 300-1 to sensor 300-n for each of sensor server 200-1 to sensor server 200-k based on the acquired information specifying the sensor in which a movement different from that in normal times occurs. From the time-series data of sensor data corresponding to each of get.
Based on the time-series data of the sensor data corresponding to each of the acquired sensors 300-1 to 300-n and the timing at which movement different from normal time is detected, the failure level determination AI detects the sensor for several hours before and after that timing. Based on the time-series data of the sensor data corresponding to each of the sensors 300-1 to 300-n, the failure level is determined (3).
The abnormality determination AI determines whether or not to issue an abnormality report for notifying that an abnormality has occurred based on the determination result of the failure level. When the monitoring device 100 c determines to issue an anomaly report, the monitoring device 100 c creates an anomaly occurrence notification and transmits the created anomaly occurrence notification to the web server 400 .
The monitoring device 100c included in the monitoring system 1c will be described in detail below.

11 is a diagram illustrating an example of a monitoring device included in a monitoring system according to Modification 3 of the embodiment; FIG.
(monitoring device 100c)
The monitoring device 100a is implemented by a device such as a personal computer, server, smart phone, tablet computer, or industrial computer. An example of the monitoring device 100c is installed at a location away from power receiving and transforming equipment such as a cubicle (high voltage power receiving equipment). The monitoring device 100c includes a communication unit 110, a creation unit 120c, a first processing unit 130c-1, a second processing unit 130c-2, an output unit 140c, and a storage unit 150, for example.
Creation unit 120c acquires sensor data corresponding to each of sensors 300-1 to 300-n received by communication unit 110 from each of sensor servers 200-1 to 200-k. Creation unit 120c creates sensor data addressed to web server 400, including sensor data corresponding to each of sensors 300-1 to 300-n acquired for each of sensor servers 200-1 to 200-k. Create notifications. The creation unit 120 c outputs the created sensor data notification to the communication unit 110 .

The first processing unit 130b-1 can be applied to the first processing unit 130c-1. The sensor data DB 132c-1 can apply the sensor data DB 132b-1.
The second processing unit 130c-2 has a learning model 132c-2. The learning model 132c-2 includes sensor data corresponding to each of the sensors 300-1 to 300-n and sensor data corresponding to each of the sensors 300-1 to 300-n for each of the sensor servers 200-1 to 200-k. sensor data corresponding to at least a portion of each of the sensors 300-1 through 300-n and the failure level of the object on which the sensors 300-1 through 300-n are armed, based on the failure level of the armed object; Obtained by machine learning the relationship with the information that identifies the level.
An example of machine learning is supervised learning. For example, a decision tree approach may be used for machine learning. A decision tree is a model that uses a tree structure to perform classification and regression. Values such as numerical values and classes are output from the root node, which is the origin, through conditional branching to the leaf node at the tip. Examples of failure levels are no failure, minor failure, medium failure, and major failure.
For each of the sensor servers 200-1 to 200-k, based on the sensor data corresponding to each of the sensors 300-1 to 300-n, the object equipped with the sensors 300-1 to 300-n is determined. The relationship between sensor data corresponding to each of the sensors 300-1 to 300-n and information specifying the failure level of the object on which the sensors 300-1 to 300-n are provided. The learning model 132c-2 may be updated by machine-learning the .

The second processing unit 130c-2 acquires sensor data corresponding to each of the sensors 300-1 to 300-n received by the communication unit 110 from each of the sensor servers 200-1 to 200-k. The second processing unit 130c-2 acquires information specifying the timing at which the sensor data output by the first processing unit 130c-1 and determined to be abnormal was acquired, and information specifying the sensor corresponding to the sensor data. do.
The second processing unit 130c-2 corresponds to each of the sensors 300-1 to 300-n acquired for each of the sensor servers 200-1 to 200-k based on the information specifying the acquired sensors. From the time-series data of the sensor data, time-series data of the sensor data corresponding to each of the sensors 300-1 to 300-n provided in the object equipped with the sensor corresponding to the sensor data determined to be abnormal is acquired. do.
The second processing unit 130c-2, based on the information specifying the timing at which the sensor data determined to be abnormal is acquired, extracts the acquired sensor data corresponding to each of the sensors 300-1 to 300-n. The sensor data corresponding to each of the sensors 300-1 to 300-n are acquired several hours before and after the timing.
The second processing unit 130c-2 acquires sensor data corresponding to each of the sensors 300-1 to 300-n and based on the learning model 132c-2. determine the failure level of the target.

When the second processing unit 130c-2 determines that the failure level of the target provided with the sensors 300-1 to 300-n is any one of a minor failure, a medium failure, and a major failure, an abnormality has occurred. In order to notify that an abnormality has occurred when it is determined that an abnormality report is to be issued to notify that an abnormality has occurred, and that the target failure level provided with the sensors 300-1 to 300-n is no failure. is determined not to issue an abnormal alarm.
When the second processing unit 130c-2 determines to issue an abnormality report, the creating unit 120c includes information specifying that an abnormality has occurred and information specifying a target failure level. Create an anomaly notification addressed to . Creation unit 120 c outputs the created abnormality occurrence notification to communication unit 110 .
The output unit 140c acquires the target failure level determination result from the second processing unit 130c-2, and outputs the acquired target failure level determination result. For example, the output unit 140c may output the determination result of the target failure level by displaying it on a display unit (not shown) or outputting it by voice.

The creation unit 120c, the first processing unit 130c-1, the second processing unit 130c-2, and the output unit 140c are computer programs (software) stored in the storage unit 150 by a hardware processor such as a CPU. It is realized by executing
Also, some or all of these functional units may be implemented by hardware (including circuit units) such as LSI, ASIC, FPGA, GPU, etc., or by cooperation between software and hardware. may be implemented. The computer program may be stored in advance in a storage device such as an HDD or flash memory, or may be stored in a removable storage medium such as a DVD or CD-ROM, and the storage medium may be loaded into a drive device. may be installed with

(Operation of monitoring system)
FIG. 12 is a flowchart illustrating an example of the operation of the monitoring system according to Modification 3 of the embodiment. Here, as an example, in the monitoring device 100c, the first processing unit 130c-1 includes sensors corresponding to the sensors 300-1 to 300-n for each of the sensor servers 200-1 to 200-k. Based on the time-series data of the data, unsupervised learning is performed on the environment in which the targets for determining anomalies are provided. Specifically, for each of the sensor servers 200-1 to 200-k, sensor data corresponding to each of the sensors 300-1 to 300-n and data corresponding to each of the sensors 300-1 to 300-n A case will be described where the relationship between patterns of corresponding sensor data is learned.
Further, the second processing unit 130c-2 generates sensor data corresponding to at least a part of each of the sensors 300-1 to 300-n for each of the sensor servers 200-1 to 200-k, and sensor data corresponding to at least a portion of each of sensors 300-1 through 300-n and sensor data corresponding to at least a portion of each of sensors 300-1 through 300-n based on the failure level of the object on which sensors 300-1 through 300-n are equipped; A case will be described in which a learning model 132c-2 obtained by machine-learning a relationship with information specifying a target failure level provided with n is provided.
Steps S1-4 to S13-4 can be applied to steps S1-3 to S13-3 described with reference to FIG. 10, so description thereof will be omitted here.

(Step S14-4)
In the monitoring device 100c, the second processing unit 130c-2 receives sensor data corresponding to each of the sensors 300-1 to 300-n from the sensor servers 200-1 to 200-k. to get The second processing unit 130c-2 acquires information specifying the timing at which the sensor data output by the first processing unit 130c-1 and determined to be abnormal was acquired, and information specifying the sensor corresponding to the sensor data. do.
The second processing unit 130c-2 corresponds to each of the sensors 300-1 to 300-n acquired for each of the sensor servers 200-1 to 200-k based on the information specifying the acquired sensors. From the time-series data of the sensor data, time-series data of the sensor data corresponding to each of the sensors 300-1 to 300-n provided in the object equipped with the sensor corresponding to the sensor data determined to be abnormal is acquired. do.
The second processing unit 130c-2 extracts the acquired sensor data corresponding to each of the sensors 300-1 to 300-n based on the information specifying the acquisition timing of the acquired sensor data determined to be abnormal. , sensor data corresponding to each of the sensors 300-1 to 300-n are acquired several hours before and after the timing.
The second processing unit 130c-2 acquires sensor data corresponding to each of the sensors 300-1 to 300-n and based on the learning model 132c-2. determine the failure level of the target.
(Step S15-4)
In the monitoring device 100c, the second processing unit 130c-2 determines whether or not to issue an anomaly report based on the determination results of the failure levels of the targets provided with the sensors 300-1 to 300-n. If it is determined not to issue an anomaly report, the process returns to step S9-4.

(Step S16-4)
In the monitoring device 100c, when the second processing unit 130c-2 determines to issue an anomaly report, the creating unit 120c creates information specifying that an anomaly has occurred and information identifying a target in which an anomaly has occurred. , and information specifying the failure level of interest, and an abnormality occurrence notification addressed to the web server 400 is created.
(Step S17-4)
In the monitoring device 100 c , the creation unit 120 c outputs the created abnormality notification to the communication unit 110 . The communication unit 110 acquires the abnormality occurrence notification output by the creation unit 120 c and transmits the acquired abnormality occurrence notification to the web server 400 .
The web server 400 receives the abnormality notification sent by the monitoring device 100c. Web server 400 creates an object such as an image for notifying that an abnormality has occurred, based on the received abnormality occurrence notification.
The terminal device 500 acquires an image for notifying the state of the monitoring target from the web server 400, and displays the acquired image for notifying the state of the monitoring target. .
In the flowchart shown in FIG. 12, the monitoring device 100c may perform the above-described processing based on sensor data transmitted by at least one of the sensor servers 200-1 to 200-k.

In the third modification of the above-described embodiment, the second processing unit 130c-2 generates sensor data corresponding to at least a part of each of the sensors 300-1 to 300-n and the sensor data from the sensors 300-1 to 300-n. sensor data corresponding to at least a part of each of the sensors 300-1 to 300-n and the sensor 300 -1 to the learning model 132c-2 obtained by machine learning the relationship with information specifying no failure, minor failure, moderate failure, and major failure as the target failure level provided with the sensor 300-n Although the case has been described, it is not limited to this example.
For example, one to three failure levels may be applied, or five or more failure levels may be applied. Also, the failure level may or may not include no failure.

In the third modification of the above-described embodiment, the second processing unit 130c-2 uses the sensor data corresponding to each of the sensors 300-1 to 300-n and the learning model 132c-2 to determine the sensor 300- 1 to sensor 300-n have been described, but the present invention is not limited to this example.
For example, the second processing unit 130c-2, based on the sensor data corresponding to at least part of the acquired sensor data corresponding to each of the sensors 300-1 to 300-n and the learning model 132c-2, A fault level may be determined for the objects on which sensors 300-1 through 300-n are provided.
Further, for example, the second processing unit 130c-2 selects sensor data other than the sensor data corresponding to at least a part of each of the sensors 300-1 to 300-n determined to be abnormal. Based on the sensors and the learning model 132c-2, the failure level of the object equipped with sensors 300-1 through 300-n may be determined.
In Modification 3 of the above-described embodiment, the monitoring device 100c may perform the above-described processing based on sensor data transmitted by at least one of the sensor servers 200-1 to 200-k.
In the third modified example of the above-described embodiment, the monitoring device 100c may include one or more second sensors in addition to the sensors 300-1 to 300-n provided for the abnormality determination targets.
Second processing unit 130c-2 further based on sensor data corresponding to at least part of each of the one or more second sensors and learning model 132c-2, sensor 300-1 to sensor 300-n A failure level of the provisioned object may be determined.
In this case, the learning model 132c-2 includes sensor data corresponding to at least part of each of the sensors 300-1 to 300-n and sensor data corresponding to at least part of each of the one or more second sensors. , and an abnormality occurring in the object on which the sensors 300-1 to 300-n are provided, sensor data corresponding to at least a part of each of the sensors 300-1 to 300-n and one or more second sensors. It is obtained by machine learning the relationship between sensor data corresponding to at least a part of each of the two sensors and information specifying the failure level of the target provided with the sensors 300-1 to 300-n. There may be.

According to the monitoring system 1c according to Modification 3 of the embodiment, in the monitoring system 1b, the second processing unit 130c-2 controls each of one or a plurality of sensors provided for each of the plurality of targets. learns the relationship between one or a plurality of sensor data combinations and the failure level of each of the plurality of devices included in the abnormality determination target, based on the sensor data combinations detected by .
With this configuration, the second processing unit 130c-2 combines sensor data detected by each of the one or more sensors 300-1 to 300-n provided for each of the one or more targets. and the failure level of the object on which the sensors 300-1 to 300-n are provided, the combination of the sensor data corresponding to each of the sensors 300-1 to 300-n and the sensor 300-1 to the sensor 300-n can be machine-learned in relation to information specifying the target failure level.
Therefore, based on the sensor data detected by each of the one or more sensors 300-1 to 300-n, it is possible to determine whether or not to issue an anomaly report.
In addition, since it is possible to improve the efficiency of on-site work and incorporate the knowledge of skilled engineers, it is possible to standardize the detection of signs of failure by AI. Combining sensors and AI can replace the five senses of engineers. Attach a sensor to the power receiving and transforming equipment of a private electric facility, detect changes in measured values, and make AI learn along with the judgment result of the failure level. In addition, it is possible to collate the results of inspections and measurements performed by engineers with the judgment results of AI to improve the evaluation and accuracy of AI.
In the monitoring system 1c, the second processing unit 130c-2 determines an abnormality based on sensor data detected by one or more sensors provided in each of the plurality of targets. update the relationship between the combination of the plurality of sensor data and the failure level of each of the plurality of devices included in the abnormality determination target based on the failure level of each of the devices.
By configuring in this way, it is possible to update the relationship between the combination of one or more sensor data and the failure level of each of the plurality of devices included in the abnormality determination target. It is possible to improve the judgment accuracy.

Although embodiments of the present invention have been described above, these embodiments are presented as examples and are not intended to limit the scope of the invention. Embodiments can be implemented in various other forms, and various omissions, replacements, and modifications can be made without departing from the scope of the invention. Embodiments and modifications thereof include, for example, those that can be easily imagined by those skilled in the art, those that are substantially the same, and those within an equivalent range.
For example, a computer program for realizing the functions of the devices described above may be recorded on a computer-readable recording medium, and the computer program recorded on this recording medium may be read and executed by a computer system. . Note that the “computer system” referred to here may include hardware such as an OS and peripheral devices.
In addition, "computer-readable recording medium" means writable non-volatile memory such as flexible disk, magneto-optical disk, ROM, flash memory, etc., portable medium such as DVD (Digital Versatile Disc), built-in computer system A storage device such as a hard disk that
Furthermore, "computer-readable recording medium" means a volatile memory (e.g., DRAM ( It also includes those that hold programs for a certain period of time, such as Dynamic Random Access Memory)).
Further, the above program may be transmitted from a computer system storing this program in a storage device or the like to another computer system via a transmission medium or by a transmission wave in a transmission medium. Here, the "transmission medium" for transmitting the program refers to a medium having a function of transmitting information, such as a network (communication network) such as the Internet or a communication line (communication line) such as a telephone line.
Further, the program may be for realizing part of the functions described above.
Further, it may be a so-called difference file (difference program) that can realize the above-described functions in combination with a program already recorded in the computer system.

1, 1a... Monitoring system 100, 100a, 100b, 100c... Monitoring device 110... Communication unit 120, 120a, 120b, 120c... Creation unit 130-1, 130b-1, 130c-1... First processing unit , 132-1, 132b-1, 132c-1c... sensor data DB, 130-2, 130a-2, 130b-2, 130c-2... second processing unit, 132-2, 132a-2, 132b-2, 132c-2... learning model, 140, 140a, 140b, 140c... output unit, 150... storage unit, 200... sensor server, 210... gateway, 220... sensor box, 230... conversion box, 300-1,..., 300-n... sensor, 400... web server, 500... terminal device

Claims (16)

A monitoring system for monitoring any of substation facilities, plants, factories, and buildings,
a plurality of sensors provided on a target for monitoring an abnormality;
determining for each sensor whether or not the time-series data of the sensor data constantly detected by each of the plurality of sensors provided for the object behaves differently than in normal times, and selecting one of the plurality of sensor data; When it is determined that the time-series data of the part of the sensor data is moving differently than in normal times, the part of the sensor data determined to be moving differently than in normal times is specified, and the sensor data is acquired. a first processing unit that acquires the timing of
When the first processing unit determines that the time-series data of a part of the sensor data out of the plurality of sensor data is moving differently than in normal times, it is determined that it is moving differently than in normal times. Acquisition of timing when a part of the sensor data is acquired , acquisition of sensor data detected by each of the plurality of sensors based on the acquired timing , and alarming of an abnormality based on the plurality of acquired sensor data A second processing unit that determines whether to
The second processing unit learns a relationship between sensor data detected by each of the plurality of sensors provided for each of the plurality of targets and an abnormality related to the target, and uses the learning result to 7. A monitoring system for determining whether or not to issue an abnormality report for a target having a sensor that outputs said part of sensor data determined by said first processing unit to be moving differently than normal. The first processing unit detects sensor data corresponding to each of the plurality of sensors based on time-series data of sensor data detected by each of the plurality of sensors provided for each of the plurality of targets. When unsupervised learning of data patterns is performed, and the sensor data detected by each of the plurality of sensors corresponds to each of the plurality of sensors that have already been acquired, using the learning results 2. The monitoring system according to claim 1, which determines whether or not it can be classified into series data, and determines that it behaves differently from normal when it cannot be classified. A monitoring system for monitoring any of substation facilities, plants, factories, and buildings,
a plurality of sensors provided on a target for monitoring an abnormality;
determining for each sensor whether or not the time-series data of the sensor data constantly detected by each of the plurality of sensors provided for the object behaves differently than in normal times, and selecting one of the plurality of sensor data; When it is determined that the time-series data of the part of the sensor data is moving differently than in normal times, the part of the sensor data determined to be moving differently than in normal times is specified, and the sensor data is acquired. a first processing unit that acquires the timing of
When the first processing unit determines that the time-series data of a part of the sensor data of the plurality of sensor data provided for the object is moving differently from normal times, the movement is different from normal times. acquiring the timing of acquiring the part of the sensor data determined to be present, acquiring the sensor data detected by each of the plurality of sensors based on the acquired timing, and acquiring the sensor data detected by each of the plurality of sensors based on the acquired timing; A second processing unit that determines whether to issue an abnormality report based on the detected sensor data,
The first processing unit detects sensor data corresponding to each of the plurality of sensors based on time-series data of sensor data detected by each of the plurality of sensors provided for each of the plurality of targets. When unsupervised learning of data patterns is performed, and the sensor data detected by each of the plurality of sensors corresponds to each of the plurality of sensors that have already been acquired, using the learning results Determine whether or not it can be classified into series data, and if it cannot be classified, determine that it is moving differently than normal,
The second processing unit learns a relationship between sensor data detected by each of the plurality of sensors provided for each of the plurality of targets and an abnormality related to the target, and uses the learning result to , a monitoring system for determining whether or not to issue an abnormality report for a target provided with a sensor that outputs said part of the sensor data determined by said first processing unit to be moving differently than normal. 4. The first processing unit according to any one of claims 1 to 3, wherein after learning the sensor data detected by each of the plurality of sensors, the first processing unit preferentially uses the most recent sensor data to re-learn. The surveillance system described in . The second processing unit detects the sensor data of the part determined to be moving differently from normal times, and the plurality of sensors provided to the target having the part of the sensor detects each of the sensors. Based on the sensor data detected by the sensors other than the part of the sensor data determined to be moving differently from normal among the detected sensor data, an abnormality alarm is issued based on the sensor data detected. The monitoring system according to any one of claims 1 to 4, which determines whether or not. The first processing unit determines an environment in which each of the plurality of sensor data and the abnormality is determined based on sensor data detected by each of the plurality of sensors provided on the target. 6. A monitoring system as claimed in any one of claims 1 to 5, wherein the system learns a relationship with. The second processing unit determines the state of each of the plurality of devices included in the targets for abnormality determination based on sensor data detected by each of the plurality of sensors provided in each of the plurality of targets. The relationship between each of the plurality of sensor data and the abnormality of each of the plurality of devices included in the target for abnormality determination is updated based on A surveillance system as described. The second processing unit determines a combination of a plurality of sensor data and an abnormality based on a combination of sensor data detected by each of the plurality of sensors provided for each of the plurality of targets. 8. The monitoring system according to any one of claims 1 to 7, wherein the relationship with each abnormality of a plurality of devices included in the object is learned. The second processing unit determines a combination of a plurality of sensor data and an abnormality based on a combination of sensor data detected by each of the plurality of sensors provided for each of the plurality of targets. 9. The monitoring system according to any one of claims 1 to 8, wherein the relationship with each failure level of a plurality of devices included in the object is learned. The second processing unit determines a failure of each of the plurality of devices included in the target for determining abnormality based on sensor data detected by each of the plurality of sensors provided in each of the plurality of targets. 10. The monitoring system according to claim 9, wherein, based on the level, the relationship between the combination of the plurality of sensor data and the failure level of each of the plurality of devices included in the object for determining abnormality is updated. 11. The monitoring system according to any one of claims 1 to 10, wherein each of said one or more sensors is one of an ultrasonic sensor, an ozone sensor, an odor sensor and a sound wave sensor. Further comprising one or more second sensors other than the plurality of sensors provided in the object for determining abnormality,
12. The second processing unit according to any one of claims 1 to 11, further based on sensor data detected by each of the one or more second sensors, determining whether or not to report an abnormality. The surveillance system described in . A step of determining whether or not time-series data of sensor data constantly detected by each of a plurality of sensors provided in the target to monitor the target for abnormality determination is different from normal behavior for each of the sensors. When,
When it is determined in the determining step that the time-series data of a part of the sensor data out of the plurality of sensor data is moving differently than in normal times, the one determined to be moving differently than in normal times Identifying the sensor data of the part and acquiring the timing at which the sensor data was acquired;
Acquiring the sensor data detected by each of the plurality of sensors based on the timing of acquiring the part of the sensor data, and determining whether or not to issue an abnormality alarm based on the acquired plurality of sensor data and
In the step of determining whether or not to issue an anomaly report, learning results of a relationship between sensor data detected by each of the plurality of sensors provided for each of the plurality of targets and anomalies related to the targets are used. , determining whether or not to issue an anomaly report for an object equipped with a sensor that outputs the part of the sensor data determined to be moving differently than normal, power receiving and transforming equipment, plant, factory, building A monitoring method performed by a computer in a monitoring system that monitors any of the A step of determining whether or not time-series data of sensor data constantly detected by each of a plurality of sensors provided in the target to monitor the target for abnormality determination is different from normal behavior for each of the sensors. When,
When it is determined in the determining step that the time-series data of a part of the plurality of sensor data provided for the object is moving differently from normal times, the movement is different from normal times. Identifying the part of the sensor data determined as and acquiring the timing at which the sensor data was acquired;
Acquiring the sensor data detected by each of the plurality of sensors based on the timing of acquiring the part of the sensor data, and determining whether or not to issue an abnormality alarm based on the acquired plurality of sensor data and
In the step of determining for each sensor, sensor data corresponding to each of the plurality of sensors is determined based on time-series data of sensor data detected by each of the plurality of sensors provided for each of the plurality of targets. , unsupervised learning of patterns between sensor data, and using the learning result, the sensor data detected by each of the plurality of sensors corresponds to each of the plurality of sensors already acquired. If it cannot be classified, it is judged that it is moving differently than normal,
In the step of determining whether or not to issue an anomaly report, the relationship between sensor data detected by each of the plurality of sensors provided for each of the plurality of targets and the anomaly related to the target is learned. Using the result of learning, whether or not to issue an anomaly report for the target equipped with the sensor that outputs the part of the sensor data that is determined to be moving differently from normal in the step of determining for each sensor A computer-executed monitoring method of a monitoring system for monitoring any one of substations, plants, factories, and buildings. In the computer of the monitoring system that monitors any of the substation facilities, plants, factories, and buildings,
A step of determining whether or not time-series data of sensor data constantly detected by each of a plurality of sensors provided in the target to monitor the target for abnormality determination is different from normal behavior for each of the sensors. When,
When it is determined in the determining step that the time-series data of a part of the sensor data out of the plurality of sensor data is moving differently than in normal times, the one determined to be moving differently than in normal times Identifying the sensor data of the part and acquiring the timing at which the sensor data was acquired;
Acquiring the sensor data detected by each of the plurality of sensors based on the timing of acquiring the part of the sensor data, and determining whether or not to issue an abnormality alarm based on the acquired plurality of sensor data , and
In the step of determining whether or not to issue an anomaly report, learning results of a relationship between sensor data detected by each of the plurality of sensors provided for each of the plurality of targets and anomalies related to the targets are used. 2. A monitoring program for determining whether or not to report an abnormality to a target having a sensor that outputs the part of the sensor data determined to be moving differently than normal. In the computer of the monitoring system that monitors any of the substation facilities, plants, factories, and buildings,
A step of determining whether or not time-series data of sensor data constantly detected by each of a plurality of sensors provided in the target to monitor the target for abnormality determination is different from normal behavior for each of the sensors. When,
When it is determined in the determining step that the time-series data of a part of the plurality of sensor data provided for the object is moving differently from normal times, the movement is different from normal times. Identifying the part of the sensor data determined as and acquiring the timing at which the sensor data was acquired;
Acquiring the sensor data detected by each of the plurality of sensors based on the timing of acquiring the part of the sensor data, and determining whether or not to issue an abnormality alarm based on the acquired plurality of sensor data , and
In the step of determining for each sensor, sensor data corresponding to each of the plurality of sensors is determined based on time-series data of sensor data detected by each of the plurality of sensors provided for each of the plurality of targets. , unsupervised learning of patterns between sensor data, and using the learning result, the sensor data detected by each of the plurality of sensors corresponds to each of the plurality of sensors already acquired. Determine whether or not it can be classified into time-series data, and if it cannot be classified, determine that it is moving differently than normal,
In the step of determining whether or not to issue an anomaly report, the relationship between sensor data detected by each of the plurality of sensors provided for each of the plurality of targets and the anomaly related to the plurality of targets is learned. Using the result of learning, whether or not to issue an anomaly report for a target equipped with a sensor that outputs some of the sensor data determined to be moving differently than normal in the step of determining for each sensor A monitoring program that determines

Images