Pseudo-abnormal data generation device, equipment monitoring system, pseudo-abnormal data generation method, and program

Abstract

This pseudo-abnormal data generation device comprises: a group generation unit that generates, from a plurality of abnormal data, each of which is time-series data, and on the basis of at least feature amounts of the plurality of abnormal data, a group including at least some of the plurality of abnormal data; a band generation unit that generates, at each time, a band indicating a range of values including the abnormal data included in the group generated by the group generation unit; and a band modulation unit that generates pseudo-abnormal data on the basis of the bands generated by the band generation unit. Thereby, pseudo-abnormal data for carrying out simulations and experiments can be generated without any expertise.

Description

Pseudo-abnormal data generating device, facility monitoring system, pseudo-abnormal data generating method, and program

[0001] The present invention relates to a pseudo-abnormal data generating device, a facility monitoring system, a pseudo-abnormal data generating method, and a program.

[0002] Learning models for detecting abnormalities in facility equipment are generated by machine learning. However, because the amount of data required for abnormalities is small, a technique for generating pseudo-abnormal data for generating learning models for detecting abnormalities has been disclosed (see Patent Literature 1). This pseudo-abnormal data is generated using simulations based on a physical model of the facility, experiments in which components are intentionally degraded or broken, and so on.

[0003] Japanese Patent Application Publication No. 2019-133212

[0004] However, the above-described method for generating pseudo-abnormal data has a problem in that it may require know-how for carrying out simulations or experiments.

[0005] The present invention has been made in consideration of the above circumstances, and provides a pseudo-abnormal data generation device, a facility monitoring system, a pseudo-abnormal data generation method, and a program for generating pseudo-abnormal data without using know-how for conducting simulations and experiments.

[0006] One aspect of the present invention is a pseudo-abnormal data generation device that includes a group generation unit that generates groups containing at least some of a plurality of abnormal data sets, each of which is time-series data, based on at least feature quantities of the plurality of abnormal data sets; a band generation unit that generates, at each time point, bands that indicate ranges of values ​​that include the abnormal data sets contained in the groups generated by the group generation unit; and a band modulation unit that generates pseudo-abnormal data based on the bands generated by the band generation unit.

[0007] Another aspect of the present invention is the pseudo-abnormal data generation device described above, wherein the band generation unit aligns the positions of the abnormal data included in the group based on the features of each of the abnormal data included in the group, and generates the band that includes the aligned abnormal data.

[0008] Another aspect of the present invention is the pseudo-abnormal data generating device described above, wherein the band modulation unit generates the pseudo-abnormal data by using the maximum or minimum value of the range of values ​​indicated by the band generated by the band generation unit as a reference and making changes to the maximum or minimum value.

[0009] Another aspect of the present invention is the pseudo-abnormal data generating device described above, wherein the band modulation section modulates the maximum or minimum value for a portion of the time of the band.

[0010] Another aspect of the present invention is an equipment monitoring system including: a group generation unit that generates, from a plurality of abnormal data sets, each of which is time-series data, groups including at least a portion of the plurality of abnormal data sets based on at least feature quantities of the plurality of abnormal data sets; a band generation unit that generates, at each time point, bands indicating ranges of values ​​that include the abnormal data sets included in the groups generated by the group generation unit; and a band modulation unit that generates pseudo-abnormal data based on the bands generated by the band generation unit.

[0011] Another aspect of the present invention is the above-mentioned equipment monitoring system, further comprising a learning model generation unit that uses the pseudo-abnormal data to generate a learning model for detecting abnormalities or signs of abnormalities from operation data, and an abnormality sign detection unit that uses the learning model generated by the learning model generation unit to detect abnormalities or signs of abnormalities from the operation data.

[0012] Another aspect of the present invention is the facility monitoring system described above, further comprising a sensor unit that measures operation data including the abnormality data.

[0013] Another aspect of the present invention is the above-mentioned equipment monitoring system, comprising a sensor unit that measures operation data including the abnormality data, and a data collection unit that collects the operation data measured by the sensor unit.

[0014] Another aspect of the present invention is a pseudo-abnormal data generation method including: a first step of generating groups including at least some of a plurality of abnormal data sets, each of which is time-series data, based on at least feature quantities of the plurality of abnormal data sets; a second step of generating, at each time point, bands indicating ranges of values ​​that include the abnormal data sets included in the groups generated in the first step; and a third step of generating pseudo-abnormal data sets based on the bands generated in the second step.

[0015] Another aspect of the present invention is a program for causing a program to function as a group generation unit that generates groups containing at least a portion of a plurality of abnormal data sets, each of which is time-series data, based on at least feature quantities of the plurality of abnormal data sets; a band generation unit that generates, at each time point, bands indicating a range of values ​​that includes the abnormal data sets contained in the groups generated by the group generation unit; and a band modulation unit that generates pseudo-abnormal data based on the bands generated by the band generation unit.

[0016] The pseudo abnormal data generating device of the present invention can generate pseudo abnormal data without using know-how for conducting simulations and experiments.

[0017] FIG. 1 is a schematic block diagram showing the configuration of a power transmission and distribution grid monitoring system 10 according to a first embodiment of the present invention. FIG. 1 is a schematic block diagram showing the configuration of a pseudo abnormal data generator 300 according to the same embodiment. FIG. 2 is a flowchart explaining an example of operation of an abnormal waveform extraction unit 301 according to the same embodiment. FIG. 3 is a schematic diagram explaining score calculation by the abnormal waveform extraction unit 301 according to the same embodiment. FIG. 4 is a flowchart explaining an example of operation of a clustering unit 303 according to the same embodiment. FIG. 5 is a flowchart explaining an example of operation of a band generation unit 304 according to the same embodiment. FIG. 6 is a diagram explaining an example of alignment by the band generation unit 304 according to the same embodiment. FIG. 7 is a diagram explaining an example of bands generated by the band generation unit 304 according to the same embodiment. FIG. 8 is a diagram explaining an example of band modulation by a band modulation unit 305 according to the same embodiment. FIG. 9 is a diagram explaining an example of an operation screen for the band modulation unit 305 according to the same embodiment. FIG. 10 is a diagram explaining an example of operation of a waveform synthesis unit 307 according to the same embodiment. FIG. 11 is a table showing an example of output from a pseudo abnormal data output unit 306 according to the same embodiment. FIG. 12 is a schematic block diagram showing the configuration of a pseudo abnormal data generator 300 according to a second embodiment of the present invention. FIG. 13 is an explanatory diagram explaining the hardware configuration of each device according to each embodiment.

[0018] First Embodiment Hereinafter, an embodiment of the present invention will be described with reference to the drawings. Fig. 1 is a schematic block diagram showing the configuration of a power transmission and distribution network monitoring system 10 according to a first embodiment of the present invention. The power transmission and distribution network monitoring system 10 (equipment monitoring system) collects operation data measured by sensors (sensor units) installed in equipment such as repeaters A2 arranged on utility poles A1, and monitors the state of the power transmission and distribution network. The operation data measured by the sensors is time-series data, such as the voltage or current of the power distribution line at repeater A2 at regular intervals.

[0019] The power transmission and distribution grid monitoring system 10 includes a data collection server 100 (data collection unit), a learning model generation device 200 (learning model generation unit), a pseudo-abnormal data generation device 300, an abnormal sign detection device 400 (abnormal sign detection unit), and an operator terminal 500. The data collection server 100, the learning model generation device 200, the pseudo-abnormal data generation device 300, the abnormal sign detection device 400, and the operator terminal 500 are communicatively connected via a network 600 such as a local area network (LAN). The data collection server 100, the learning model generation device 200, the pseudo-abnormal data generation device 300, the abnormal sign detection device 400, and the operator terminal 500 may be realized by a computer reading and executing a program. Each of the data collection server 100, the learning model generation device 200, the pseudo-abnormal data generation device 300, the abnormal sign detection device 400, and the operator terminal 500 may be realized by a single computer or multiple computers. Furthermore, multiple devices or parts of multiple devices may be realized by a single computer.

[0020] The data collection server 100 collects and stores operation data measured by sensors. The learning model generation device 200 generates a learning model for detecting abnormal data from the operation data. After the pseudo-abnormal data generation device 300 generates the pseudo-abnormal data, the learning model generation device 200 may use the pseudo-abnormal data in addition to the operation data stored in the data collection server 100 when generating the learning model.

[0021] The pseudo abnormal data generation device 300 extracts abnormal data from the operation data using the learning model generated by the learning model generation device 200, and generates pseudo abnormal data based on the extracted abnormal data. Details of the pseudo abnormal data generation device 300 will be described later.

[0022] The abnormal sign detection device 400 detects abnormal data from the operation data using the learning model generated by the learning model generation device 200. When the abnormal sign detection device 400 detects abnormal data, it notifies the operator that there is an abnormality or a sign of an abnormality in the power transmission and distribution network via the operator terminal 500. The abnormal sign detection device 400 may determine the type of the detected abnormal data (details will be described later) and notify the operator of the type or a type of abnormality corresponding to the type via the operator terminal 500.

[0023] The operator terminal 500 is a terminal used by an operator of the power transmission and distribution network monitoring system 10, and allows the operator to configure the power transmission and distribution network monitoring system 10, and the power transmission and distribution network monitoring system 10 to notify the operator.

[0024] 2 is a schematic block diagram showing the configuration of a pseudo abnormal data generation device 300 according to this embodiment. The pseudo abnormal data generation device 300 includes an abnormal waveform extraction unit 301, a type classification unit 302, a clustering unit 303, a band generation unit 304, a band modulation unit 305, a pseudo abnormal data output unit 306, and a waveform synthesis unit 307. The type classification unit 302 and the clustering unit 303 may form a group generation unit.

[0025] The abnormal waveform extraction unit 301 uses the learning model generated by the learning model generation device 200 to extract abnormal data (abnormal waveforms) from the operation data collected by the data collection server 100 .

[0026] The type classification unit 302 classifies each piece of abnormal data extracted by the abnormal waveform extraction unit 301 into multiple types based on its feature values. The feature values ​​for each piece of abnormal data may include at least the average value of the abnormal data, the median value of the abnormal data, the number of times the abnormal data crosses the average value or the median value, the difference between the maximum and minimum values ​​of the abnormal data, the absolute value of the difference between the values ​​at the start and end points of the abnormal data, and the slope or sign of the abnormal data at the first crossing point of the abnormal data with the average value or the median value. In this embodiment, the type classification unit 302 classifies the abnormal data into six types: upper peak type, lower peak type, upper and lower peak type, transient rise type, transient fall type, and oscillation type, using the method described in Japanese Patent No. 6827608.

[0027] The clustering unit 303 further classifies the abnormal data of each type classified by the type classification unit 302 into a plurality of clusters (groups). For example, the clustering unit 303 performs classification using the K-means method based on a feature such as the difference between the maximum value and the minimum value or the distance between abnormal data such as the Euclidean distance.

[0028] For each group generated by the clustering unit 303, the band generating unit 304 generates a band indicating a range of values ​​that includes the abnormal data included in that group at each time. Note that before generating the bands, the band generating unit 304 may align the abnormal data in the time direction or in the magnitude direction based on the feature of the abnormal data. The alignment based on the feature may be, for example, aligning the time of the maximum value, the time of the average value, or the magnitude of the average value.

[0029] The band modulation unit 305 generates pseudo-abnormal data based on each band generated by the band generation unit 304. For example, the band modulation unit 305 generates pseudo-abnormal data by subtracting a random value from some or all of the time-series data listing the maximum values ​​of the bands at each time, or by multiplying the data by a random percentage. For example, the pseudo-abnormal data at time t may be expressed as D(t) = Max(t) - R(t), where Max(t) is the maximum value, Min(t) is the minimum value, and R(t) is the random value, respectively, of the band at time t. Alternatively, the pseudo-abnormal data D(t) may be expressed as D(t) = Min(t) + R(t) × (D(t) - Min(t)), where 0≦R(t)≦1.

[0030] The pseudo-abnormal data output unit 306 outputs the pseudo-abnormal data generated by the band modulation unit 305 and the waveform synthesis unit 307. Note that the pseudo-abnormal data output unit 306 may output only the pseudo-abnormal data that has been visually confirmed by the operator, for example, by displaying a graph of the pseudo-abnormal data generated by the band modulation unit 305 and the waveform synthesis unit 307 on the operator terminal 500.

[0031] The waveform synthesis unit 307 synthesizes the abnormal data with waveform data corresponding to the type classified by the type classification unit 302, to generate pseudo-abnormal data. Which waveform data to synthesize for each type may be set in advance by an operator via the operator terminal 500, or may be set when the pseudo-abnormal data generation device 300 is manufactured.

[0032] 3 is a flowchart illustrating an example of the operation of the abnormal waveform extraction unit 301 in this embodiment. First, the abnormal waveform extraction unit 301 divides the operation data into windows of a predetermined time length (step Sa1). This predetermined time length may be set in advance by an operator via the operator terminal 500, or may be set when the pseudo abnormal data generation device 300 is manufactured.

[0033] Next, the abnormal waveform extraction unit 301 calculates the distance between each window (step Sa2). Here, the distance is a value indicating the similarity between the time series data constituting the windows, and the more similar the distance, the smaller the value. For example, the distance can be calculated using Discord (Takaaki Nakamura et al., "Time Series Data Singularity Detection Method Using Sample Time Series Extraction," DEIM Forum 2015 F8-2, https: / / db-event.jpn.org / deim2015 / paper / 16.pdf).

[0034] Next, the abnormal waveform extraction unit 301 calculates a score for each window (step Sa3). This score is a value that increases if there are no windows similar to the time-series data that constitutes that window. For example, the score is set to the smallest value among the distances between that window and other windows.

[0035] Next, the abnormal waveform extraction unit 301 obtains a score threshold from the learning model generation device 200 (step Sa4). This score threshold is the learning model generated by the learning model generation device 200. As in steps Sa1 to Sa3, the learning model generation device 200 calculates a score for each window of operational data prepared for generating the learning model, and determines the score threshold based on the distribution of the calculated scores. For example, the learning model generation device 200 may set a predetermined percentage of the top scores among the calculated scores as the score threshold, or may set a score that corresponds to a predetermined deviation value as the score threshold.

[0036] Next, the abnormal waveform extraction unit 301 extracts, as an abnormal waveform (abnormal data), a window in which the score calculated in step Sa3 exceeds the score threshold acquired in step Sa4 (step Sa5).

[0037] 4 is a schematic diagram illustrating score calculation by the abnormal waveform extraction unit 301 in this embodiment. As shown in Fig. 4, the abnormal waveform extraction unit 301 divides the operational data A into windows A1, A2, A3, A4, etc. Furthermore, for window A1, the abnormal waveform extraction unit 301 calculates a distance D12 from window A2, a distance D13 from window A3, a distance D14 from window A4, etc., and sets the smallest of these distances as the score for window A1.

[0038] 5 is a flowchart illustrating an example of the operation of the clustering unit 303 in this embodiment. The clustering unit 303 performs clustering on the abnormal data belonging to each type classified by the type classification unit 302 based on their feature amounts (step Sb1). The feature amounts used at this time may be selected according to the type. For example, the upper peak type may be clustered based on the maximum value, and the oscillation type may be clustered based on the number of intersections between the average value and the abnormal data. Alternatively, clustering may be performed based on multiple feature amounts.

[0039] Next, the clustering unit 303 clusters each cluster resulting from the clustering in step Sb1 based on the similarity of waveforms between abnormal data to generate groups of abnormal data (step Sb2). Note that when the K-means method is used for the clustering in steps Sb1 and Sb2, the value of K may be set by an operator via the operator terminal 500 for each step, or may be set when the pseudo abnormal data generation device 300 is manufactured.

[0040] 6 is a flowchart illustrating an example of the operation of the band generation unit 304 in this embodiment. The example of operation illustrated in FIG. 6 is an example of the operation of the band generation unit 304 for each group generated by the clustering unit 303. First, the band generation unit 304 selects one reference abnormal waveform from the abnormal data belonging to the group (step Sc1). This selection may be, for example, by selecting the first abnormal data belonging to the group, or by selecting the abnormal data whose feature used to generate the group is closest to the center of the group, or by specifying the reference abnormal waveform via the operator terminal 500.

[0041] Next, the band generation unit 304 aligns the other abnormal data with the reference abnormal data based on the feature (step Sc2). For example, if the group is made up of upper-peak type abnormal data, the band generation unit 304 shifts the other abnormal data in the time direction so that the time when the other abnormal data reaches its maximum value coincides with the time when the reference abnormal data reaches its maximum value. Furthermore, the band generation unit 304 shifts the other abnormal data up or down (in the direction of the abnormal data value) so that the average value of the other abnormal data coincides with the average value of the reference abnormal data.

[0042] Next, the band generator 304 generates a band that includes all of the abnormal data aligned in step Sc2 (step Sc3). For example, the band generator 304 extracts the minimum and maximum values ​​of all of the aligned abnormal data and the reference abnormal data at each time point, and defines the area enclosed by these values ​​as a band.

[0043] 7 is a diagram illustrating an example of alignment performed by the band generation unit 304 in this embodiment. The example of alignment in FIG. 7 is an example of time-direction alignment for a group consisting of upper-peak type abnormal data. As shown in FIG. 7, the band generation unit 304 shifts each of the abnormal data W2 and W3 in the time direction so that the time at which each of the abnormal data W2 and W3 reaches its maximum value is the same as the time at which the reference abnormal data W1 reaches its maximum value.

[0044] 8 is a diagram illustrating an example of bands generated by the band generation unit 304 in this embodiment. In FIG. 8, graph B1 is a graph plotting the maximum value of the aligned abnormal data at each time. Graph B2 is a graph plotting the minimum value of the aligned abnormal data at each time. The band generated by the band generation unit 304 is the region (range) sandwiched between graphs B1 and B2.

[0045] 9 is a diagram illustrating an example of band modulation by the band modulation unit 305 in this embodiment. In FIG. 9, graphs B1 and B2 are the same as graphs B1 and B2 in FIG. 8. Graph M1 is pseudo-abnormal data. In the example shown in FIG. 9, the band modulation unit 305 generates pseudo-abnormal data M1 so that it fits within the band surrounded by graphs B1 and B2.

[0046] Fig. 10 is a diagram showing an example of an operation screen for the band modulation unit 305 in this embodiment. The example of the operation screen in Fig. 10 is displayed by the band modulation unit 305 on the operator terminal 500 to allow the operator to set parameters for generating pseudo-abnormal data. In Fig. 10, input fields T1 to T9 are input fields for parameters for generating pseudo-abnormal data. An OK button T10 and a cancel button T11 are buttons that allow the operator to specify whether or not to approve the generated pseudo-abnormal data.

[0047] Input area T1 is an area for inputting a selection of whether the graph used as a reference when generating pseudo-abnormal data should be the maximum value or the minimum value of the band. Input area T2 is an area for inputting the start position (start time) of the range to be band modulated (modulation range 1). Input area T3 is an area for inputting the end position (end time) of the range to be band modulated (modulation range 1). Input area T4 is an area for inputting the maximum rate of change from the reference graph by band modulation for modulation range 1. Input area T5 is an area for inputting the frequency (rate) of time positions to be changed from the reference graph by band modulation for modulation range 1. The band modulation unit 305 randomly changes values ​​for the rate set in input area T5 among the time positions constituting the data of modulation range 1 so that the amount of change is within the rate set in input area T4 of the band width.

[0048] Input area T6 is an area for inputting the start position (start time) of the range to be band modulated (modulation range 2). Input area T7 is an area for inputting the end position (end time) of the range to be band modulated (modulation range 2). Input area T8 is an area for inputting the maximum rate of change from the reference graph by band modulation for modulation range 2. Input area T9 is an area for inputting the frequency (rate) of time positions to be changed from the reference graph by band modulation for modulation range 2. The band modulation unit 305 randomly changes the values ​​of the rate set in input area T9 among the time positions that make up the data for modulation range 2 so that the amount of change is within the rate set in input area T8 for the band width.

[0049] Graph L1 is a graph showing the boundary on the maximum value side of the band. Graph L2 is a graph showing the boundary on the minimum value side of the band. In FIG. 10 , because the maximum value is selected in input region T1, graph L1 is displayed as a solid line and graph L2 is displayed as a dashed line. Graph L3 shown in modulation range 1 is a graph obtained by changing graph L1 in modulation range 1 for the time position of the frequency set in input region T5, with the ratio set in input region T4 as the maximum value. Graph L4 shown in modulation range 2 is a graph obtained by changing graph L1 in modulation range 2 for the time position of the frequency set in input region T9, with the ratio set in input region T8 as the maximum value.

[0050] When the OK button T10 is pressed, the band modulation unit 305 sets as pseudo-abnormal data a graph consisting of graph L3 for modulation range 1, graph L4 for modulation range 2, and graph L1 for times other than those. When the cancel button T11 is pressed, the band modulation unit 305 discards the pseudo-abnormal data of that graph. Note that while Figure 10 shows an example in which there are two modulation ranges, modulation range 1 and modulation range 2, the number of modulation ranges may be one, three or more, or may be set by the operator.

[0051] 11 is a diagram illustrating an example of the operation of the waveform synthesis unit 307 in this embodiment. As shown in FIG. 11, the waveform synthesis unit 307 adds synthetic data W5 corresponding to the type of abnormal data W4 to generate pseudo-abnormal data W6. Here, "addition" refers to adding the value of the abnormal data W4 at each time to the value of the synthetic data W5. Note that the synthetic data W5 may be normal data.

[0052] Fig. 12 is a table showing an example of output from the pseudo-abnormal data output unit 306 in this embodiment. In Fig. 12, the data ID is a number that identifies the operational data or abnormal data that was the basis for generating the pseudo-abnormal data. The pseudo-abnormal data ID is a number that identifies the pseudo-abnormal data. The pseudo-abnormal data ID may be assigned in combination with the data ID to uniquely identify the pseudo-abnormal data.

[0053] The classification ID is a number indicating the type or group to which the original abnormal data belongs when generating the pseudo-abnormal data. The generation method type is a number indicating whether the pseudo-abnormal data was generated by band modulation or waveform synthesis. The time is the time corresponding to the value of the pseudo-abnormal data in the row. The pseudo-abnormal data is the value of the pseudo-abnormal data at the time in the same row.

[0054] Second Embodiment In the first embodiment, the waveform synthesis unit 307 determines the synthetic data to be synthesized with the abnormal data based on the type classified by the type classification unit 302. In the second embodiment, the synthetic data to be synthesized with the abnormal data is determined based on the cluster (group) by the clustering unit 303. The configuration of the power transmission and distribution network monitoring system 10 in this embodiment is the same as that in FIG.

[0055] 13 is a schematic block diagram showing the configuration of a pseudo abnormal data generator 300 according to a second embodiment of the present invention. In FIG. 13, parts corresponding to those in FIG. 2 are given the same reference numerals, and descriptions thereof will be omitted. As shown in FIG. 13, the pseudo abnormal data generator 300 according to this embodiment differs from the first embodiment in that the waveform synthesis unit 307 determines synthesized data based on clusters generated by the clustering unit 303.

[0056] FIG. 14 is an explanatory diagram illustrating the hardware configuration of each device according to each of the above-described embodiments. The devices are the data collection server 100, the learning model generation device 200, the pseudo-abnormal data generation device 300, the abnormal sign detection device 400, and the operator terminal 500. Each device includes an input / output module I, a storage module M, and a control module P. The input / output module I includes some or all of the following: the communication module H11, the connection module H12, the pointing device H21, the keyboard H22, the display H23, the button H3, the microphone H41, the speaker H42, the camera H51, and the sensor H52. The storage module M includes a drive H7. The storage module M may further include some or all of the memory H8. The control module P includes a memory H8 and a processor H9. These hardware components are connected to each other via a bus (Bus) and receive power from a power supply H6.

[0057] The connection module H12 is a digital input / output port such as a USB (Universal Serial Bus). In the case of a portable device, the pointing device H21, keyboard H22, and display H23 are touch panels. The sensor H52 is an acceleration sensor, a gyro sensor, a GPS receiving module, a proximity sensor, or the like. The power supply H6 is a power supply unit that supplies the electricity necessary to operate each device. In the case of a portable device, the power supply H6 is a battery. The drive H7 is an auxiliary storage medium such as a hard disk drive or a solid-state drive. The drive H7 may be a non-volatile memory such as an EEPROM or a flash memory, or a magneto-optical disk drive or a flexible disk drive. Furthermore, the drive H7 is not limited to being built into each device, but may also be an external storage device connected to the connector of the connection module H12. The memory H8 is a main storage medium such as a random access memory. The memory H8 may also be a cache memory. The memory H8 stores instructions when the instructions are executed by one or more processors H9. The processor H9 is a CPU (Central Processing Unit). The processor H9 may be an MPU (Microprocessing Unit) or a GPU (Graphics Processing Unit). The processor H9 reads programs and various data from the drive H7 via the memory H8 and performs calculations to execute instructions stored in one or more memories H8.

[0058] The input / output module I is used in the abnormal waveform extraction unit 301, the pseudo abnormal data output unit 306, the operator terminal 500, etc. The storage module M realizes the data collection server 100. The control module P is used to implement each of the data collection server 100, the learning model generation device 200, the pseudo abnormal data generation device 300, the abnormal sign detection device 400, and the operator terminal 500. Note that in this specification and the like, the terms "data collection server 100," "learning model generation device 200," "pseudo abnormal data generation device 300," "abnormal sign detection device 400," and "operator terminal 500" may be replaced with the term "control module P."

[0059] Although the above-described embodiments are examples in which the pseudo abnormal data generator 300 is used in the power transmission / distribution grid monitoring system 10 that monitors the power transmission / distribution grid, the pseudo abnormal data generator 300 may be used in systems other than the power transmission / distribution grid monitoring system 10. For example, the pseudo abnormal data generator 300 may be used in an equipment system that monitors equipment such as a factory or public facility, an observation system that observes natural phenomena such as weather, or a measurement system that measures human activity such as people flow.

[0060] Alternatively, the group generation unit may be configured to include only one of the type classification unit 302 and the clustering unit 303, and the band generation unit 304 may generate bands for the groups of abnormal data classified by the group generation unit.

[0061] The following embodiments may be possible: (1) One embodiment is a pseudo-abnormal data generation device including: a group generation unit that generates, from a plurality of abnormal data pieces each being time-series data, groups including at least a portion of the plurality of abnormal data pieces based on at least feature quantities of the plurality of abnormal data pieces; a band generation unit that generates, at each time, bands indicating ranges of values ​​that include the abnormal data pieces included in the groups generated by the group generation unit; and a band modulation unit that generates pseudo-abnormal data based on the bands generated by the band generation unit.

[0062] (2) Another embodiment is the pseudo-abnormal data generation device of (1) above, wherein the band generation unit aligns the positions of the abnormal data included in the group based on the features of each of the abnormal data included in the group, and generates the band including the aligned abnormal data.

[0063] (3) Another embodiment is the pseudo-abnormal data generating device of (1) or (2) above, wherein the band modulation unit generates the pseudo-abnormal data by using the maximum or minimum value of the range of values ​​indicated by the band generated by the band generation unit as a reference and modifying the maximum or minimum value.

[0064] (4) Another embodiment is the pseudo-abnormal data generating device of (3) above, wherein the band modulation unit modulates the maximum or minimum value for a portion of the time of the band.

[0065] (5) Another embodiment is an equipment monitoring system including: a group generation unit that generates groups including at least a portion of a plurality of abnormal data sets, each of which is time-series data, based on at least feature quantities of the plurality of abnormal data sets; a band generation unit that generates, at each time point, bands indicating ranges of values ​​that include the abnormal data sets included in the groups generated by the group generation unit; and a band modulation unit that generates pseudo-abnormal data based on the bands generated by the band generation unit.

[0066] (6) Another embodiment is the equipment monitoring system of (5) above, further comprising a learning model generation unit that uses the pseudo-abnormal data to generate a learning model for detecting an abnormality or signs of an abnormality from operation data, and an abnormality sign detection unit that uses the learning model generated by the learning model generation unit to detect an abnormality or signs of an abnormality from the operation data.

[0067] (7) Another embodiment is the facility monitoring system of (5) or (6) above, which includes a sensor unit that measures operation data including the abnormality data.

[0068] (8) Another embodiment is an equipment monitoring system according to (5) or (6) above, comprising a sensor unit that measures operation data including the abnormal data, and a data collection unit that collects the operation data measured by the sensor unit.

[0069] (9) Another embodiment is a pseudo-abnormal data generation method including: a first step of generating groups including at least some of a plurality of abnormal data sets, each of which is time-series data, based on at least feature quantities of the plurality of abnormal data sets; a second step of generating, at each time, bands indicating ranges of values ​​that include the abnormal data sets included in the groups generated in the first step; and a third step of generating pseudo-abnormal data sets based on the bands generated in the second step.

[0070] (10) Another embodiment is a program for causing a program to function as a group generation unit that generates groups including at least some of a plurality of abnormal data sets, each of which is time-series data, based on at least feature quantities of the plurality of abnormal data sets; a band generation unit that generates, at each time point, bands indicating a range of values ​​that includes the abnormal data sets included in the groups generated by the group generation unit; and a band modulation unit that generates pseudo-abnormal data based on the bands generated by the band generation unit.

[0071] 1 may be recorded on a computer-readable recording medium, and the programs recorded on this recording medium may be read into a computer system and executed to realize the data collection server 100, the learning model generation device 200, the pseudo-abnormal data generation device 300, the abnormal sign detection device 400, and the operator terminal 500. Note that the term "computer system" here includes hardware such as an OS and peripheral devices.

[0072] Furthermore, "computer system" also includes the homepage provision environment (or display environment) if the WWW system is used. Furthermore, "computer-readable recording medium" refers to portable media such as flexible disks, optical magnetic disks, ROMs, and CD-ROMs, as well as storage devices such as hard disks built into computer systems. Furthermore, "computer-readable recording medium" also includes devices that dynamically store programs for a short period of time, such as communication lines used when transmitting programs over networks like the Internet or communication lines like telephone lines, and devices that store programs for a fixed period of time, such as volatile memory within the computer systems that serve as servers or clients in such cases. Furthermore, the above-mentioned programs may be those that implement part of the aforementioned functions, or may be those that can realize the aforementioned functions in combination with programs already stored in the computer system.

[0073] Although an embodiment of the present invention has been described in detail above with reference to the drawings, the specific configuration is not limited to this embodiment, and design changes and the like are also included within the scope that does not deviate from the gist of the present invention.

[0074] 10 Power transmission and distribution network monitoring system 100 Data collection server 200 Learning model generation device 300 Pseudo abnormal data generation device 301 Abnormal waveform extraction unit 302 Type classification unit 303 Clustering unit 304 Band generation unit 305 Band modulation unit 306 Pseudo abnormal data output unit 307 Waveform synthesis unit 400 Abnormal sign detection device 500 Operator terminal 600 Network

Claims

1. A pseudo abnormal data generating device comprising:a group generating unit configured to generate a group including at least a part of a plurality of pieces of abnormal data on the basis of at least feature quantities of the plurality of pieces of abnormal data from the plurality of pieces of abnormal data that is time series data;a band generating unit configured to generate a band representing a range of values including abnormal data included in the group generated by the group generating unit at each time; anda band modulating unit configured to generate pseudo abnormal data on the basis of the band generated by the band generating unit.

2. The pseudo abnormal data generating device according to claim 1, wherein the band generating unit adjusts positions of pieces of abnormal data included in the group on the basis of feature quantities of the pieces of abnormal data included in the group and generates the band including the pieces of abnormal data of which positions have been adjusted.

3. The pseudo abnormal data generating device according to claim 1 or 2, wherein the band modulating unit generates the pseudo abnormal data by applying a change to a maximum value or a minimum value of the range of the values represented by the band generated by the band generating unit with the maximum value or the minimum value setas a reference.

4. The pseudo abnormal data generating device according to claim 3, wherein the band modulating unit performs the change applied to the maximum value or the minimum value at a time of a part of the band.

5. A facility monitoring system comprising:a group generating unit configured to generate a group including at least a part of a plurality of pieces of abnormal data on the basis of at least feature quantities of the plurality of pieces of abnormal data from the plurality of pieces of abnormal data that is time series data;a band generating unit configured to generate a band representing a range of values including abnormal data included m the group generated by the group generating unit at each time; anda band modulating unit configured to generate pseudo abnormal data on the basis of the band generated by the band generating unit.

6. The facility monitoring system according to claim 5, further comprising:a learning model generating unit configured to generate a learning model used for detecting an abnormality or a sign of an abnormality from operating data using the pseudo abnormal data; andan abnormal sign detecting unit configured to detect an abnormality or a sign of an abnormality from the operating data using the learning model generated by the learning model generating unit.

7. The facility monitoring system according to claim 5 or 6, further comprising asensor unit configured to measure operating data including the abnormal data.

8. The facility monitoring system according to claim 5 or 6, further comprisinga sensor unit configured to measure operating data including the abnormal data; anda data collecting unit configured to collect operating data measured by the sensor unit.

9. A pseudo abnormal data generating method comprising:a first step of generating a group including at least a part of a plurality of pieces of abnormal data on the basis of at least feature quantities of the plurality of pieces of abnormal data from the plurality of pieces of abnormal data that is time series data;a second step of generating a band representing a range of values including abnormal data included in the group generated in the first step at each time; anda third step of generating pseudo abnormal data on the basis of the band generated in the second step.

10. A program causing the program to function as:a group generating unit configured to generate a group including at least a part of a plurality of pieces of abnormal data on the basis of at least feature quantities of the plurality of pieces of abnormal data from the plurality of pieces of abnormal data that is time series data;a band generating unit configured to generate a band representing a range of values including abnormal data included in the group generated by the group generating unit at each time; anda band modulating unit configured to generate pseudo abnormal data on the basis of the band generated by the band generating unit.

Images