Information processing device, information processing method, and program

Abstract

To predict the probability that a second disconnector malfunctions at a scheduled operation time.SOLUTION: An information processing device including a processor and a storage device is configured to: store a learning model for conducting learning on the basis of learning data where feature amounts including data indicative of first images of the first disconnector photographed per prescribed period after installation of the first disconnector are associated with labels including data indicative of an operating period lasting from when the first disconnector is installed to when the disconnector malfunctions in the past; and predict the probability that a second disconnector malfunctions at a scheduled operation time by inputting data indicative of a second image of the second disconnector photographed after installation of the second disconnector, an elapsed period from the installation of the second disconnector, and the scheduled operation time of the second disconnector at present.SELECTED DRAWING: Figure 2

Description

[Technical field]

[0001] The present invention relates to an information processing device, an information processing method, and a program for predicting the probability that a disconnector will malfunction at a scheduled time for its operation. [Background technology]

[0002] A disconnecting switch is a switch that opens and closes an electric path of a power line in which a fault current or the like is interrupted by a circuit breaker and no current flows. The disconnecting switch is a structure that includes, for example, a fixed contactor connected to the electric path of the power line provided on the upstream side of the circuit breaker, a movable contactor that contacts the fixed contactor, and a blade that moves the movable contactor so that the movable contactor is in either a contact state or a non-contact state with the fixed contactor. After being installed in a power system, the disconnecting switch is affected by the surrounding environment of the installation site. Therefore, as the disconnecting switch ages after installation, dust adhesion, rusting, aging deterioration, and the like occur on the fixed contactor, the movable contactor, the blade, and the like. For example, as the disconnecting switch ages after installation, rust occurs on at least a part of the blade, or grease applied to smooth the operation of the blade dries, hardens, or solidifies, causing a malfunction that makes it difficult for the blade to operate, which may cause a contact failure that makes it difficult for the movable contactor to contact the fixed contactor. Furthermore, if dust or the like adheres to at least one of the movable contact and the fixed contact, the contact area between the movable contact and the fixed contact may overheat, causing a fire. Therefore, in order to prevent such problems from occurring, inspection work is carried out on each disconnecting switch installed in the power system, for example, at regular intervals (for example, every six years) after the disconnecting switch is installed (see, for example, Patent Document 1). [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2017-54593 A Summary of the Invention [Problem to be solved by the invention]

[0004] However, with the method of inspecting the disconnector at regular intervals as described above, depending on the number of times the disconnector has operated in the past and the surrounding environment of the installation location, the disconnector may already be malfunctioning before the next inspection, which could make it impossible to open or close the electrical circuit when desired.

[0005] The present invention has been made in consideration of the above-mentioned problems, and one object of the present invention is to provide an information processing device, an information processing method, and a program for predicting the probability that a disconnector will malfunction at a scheduled time for operating the disconnector (e.g., six months later, one year later, etc.). [Means for solving the problem]

[0006] One of the present inventions for achieving the above-mentioned object is an information processing device having a processor and a storage device, which stores a learning model that performs learning based on learning data in which a feature including data indicating a first image of a first disconnector taken at predetermined intervals since the installation of the first disconnector in the past is associated with a label including data indicating a period of time during which the first disconnector can be operated from the installation until it malfunctions, and currently inputs data indicating a second image of the second disconnector taken after the installation of a second disconnector, the period of time elapsed since the installation of the second disconnector, and the planned time of operation of the second disconnector into the learning model, thereby predicting the probability that the second disconnector will malfunction at the planned time of operation.

[0007] According to the information processing device of the present invention, the predicted result of the probability that the second disconnector will malfunction at the scheduled operation time can be confirmed before the scheduled operation time of the second disconnector, making it possible to perform inspection work on the second disconnector according to the probability that the second disconnector will malfunction.

[0008] In another aspect of the present invention for achieving the above-mentioned object, in an information processing device, the feature includes data indicating an installation location of the first disconnector, and currently, data indicating a second image of the second disconnector, the period of time elapsed since the second disconnector was installed, the planned time to operate the second disconnector, and the installation location of the second disconnector are input into the learning model to predict the probability that the second disconnector will malfunction at the planned time of operation.

[0009] The causes of the second disconnector malfunctioning vary depending on the environment at the installation location of the second disconnector, and the number of years from the installation of the second disconnector until it malfunctions also vary depending on the environment at the installation location of the second disconnector. According to the information processing device of the present invention, the learning model learns using a combination of features and labels including data indicating the installation location of the first disconnector, and data indicating the installation location of the second disconnector is also input to the learning model, making it possible to improve the prediction accuracy when predicting the probability that the second disconnector will malfunction at the scheduled operation time.

[0010] In another aspect of the present invention for achieving the above-mentioned object, in an information processing device, the feature includes data indicating the number of times the first disconnector operates for each specified period, and data indicating a second image of the second disconnector, the period elapsed since the second disconnector was installed, the planned time to operate the second disconnector, and the number of times the second disconnector has operated up to now are input into the learning model to predict the probability that the second disconnector will malfunction at the planned time of operation.

[0011] The operable period from the installation of the second disconnecting switch until it malfunctions tends to be shorter as the number of operations since the installation of the second disconnecting switch increases. According to the information processing device of the present invention, the learning model learns by combining the feature amount including data indicating the number of operations of the first disconnecting switch per predetermined period and the label, and data indicating the number of operations of the second disconnecting switch up to the present is also input to the learning model, so that it is possible to improve the prediction accuracy when predicting the probability that the second disconnecting switch will malfunction at the scheduled operation time.

[0012] Another aspect of the present invention for achieving the above object is, in an information processing device, the first disconnector and the second disconnector are the same type of disconnector, and the first image and the second image are images taken from the same direction relative to the first disconnector and the second disconnector, respectively.

[0013] According to the information processing device of the present invention, the dust adhesion, rust, deterioration over time, etc. occurring in the second disconnector can be easily compared with the dust adhesion, rust, deterioration over time, etc. occurring in the first disconnector, which makes it possible to improve the prediction accuracy when predicting the probability that the second disconnector will malfunction at the scheduled operation time.

[0014] Other problems and solutions disclosed in the present application will be made clear in the section of the preferred embodiment and the drawings. Effect of the Invention

[0015] According to the present invention, the probability that the second disconnecting switch will malfunction at the scheduled operation time can be predicted before the scheduled operation time of the second disconnecting switch, so that inspection work on the second disconnecting switch can be performed before the second disconnecting switch malfunctions. [Brief description of the drawings]

[0016] [Figure 1A] 1 is a perspective view showing an example of a mechanism for rotating a blade 107 of a disconnector 100, 500. FIG. [Figure 1B] FIG. 2 is a perspective view showing an example of a state in which the disconnectors 100 and 500 are in an open state. [Figure 1C] FIG. 2 is a perspective view showing an example of a state in which the disconnectors 100 and 500 are in a closed state. [Diagram 2] 1 is a block diagram showing a schematic configuration of a prediction system 1 including an information processing device 200 that predicts the probability that a disconnector 500 will malfunction at a scheduled operation time by using a learning model (AI model). [Diagram 3] FIG. 2 is a diagram showing an example of features and labels constituting training data 242. [Figure 4] 1 is a diagram showing an example of an image actually taken of a blade 107, a twisting mechanism 105, and a pivot pin 106 of a disconnector 100. FIG. [Diagram 5] A diagram showing the structure of a deep neural network, which is an example of a learning model 243. [Figure 6] 2 is a block diagram showing an example of hardware of an information processing device 200 used to realize the prediction system 1. FIG. [Figure 7] 13 is a flowchart showing a process performed by the information processing device 200 to learn a learning model 243 using learning data 242. [Figure 8] 13 is a flowchart showing the process performed by the information processing device 200 when predicting the probability that the disconnector 500 will malfunction at the scheduled operation time using the learning model 243. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0017] At least the following matters become clear from the description of this specification and the accompanying drawings. Hereinafter, the present invention will be described according to one embodiment with reference to the accompanying drawings. In this embodiment, the past disconnector (first disconnector) used to generate the learning data 242 is the disconnector 100, and the current disconnector (second disconnector) targeted for predicting the probability of malfunction at the scheduled operation time is the disconnector 500. The disconnectors 100 and 500 are the same type of disconnector, for example, horizontal two-point cut type disconnectors.

[0018] FIG. 1A is a perspective view showing an example of a mechanism for rotating the blade 107 of the disconnector 100, 500, FIG. 1B is a perspective view showing an example of the disconnector 100, 500 in an open state, and FIG. 1C is a perspective view showing an example of the disconnector 100, 500 in an on state. Since the disconnectors 100, 500 are the same type of disconnector, the elements constituting the structure of each disconnector 100, 500 are given common numbers. In addition, the disconnectors 100, 500 are arranged in parallel, for example, on the power source side upstream of the installation position of a circuit breaker (not shown), so as to be simultaneously in an open state or on state for the three-phase power lines. In addition, the disconnectors 100, 500 of each phase are structured to be simultaneously set to an open state or on state for the three-phase power lines via an interlocking mechanism (not shown) to which an operating force from the outside is transmitted. Since the disconnectors 100, 500 of each phase have the same configuration, the structure of the disconnectors 100, 500 for one phase will be described below. When describing the structure of the disconnectors 100, 500 depicted in Figures 1A to 1C, the description may be made using coordinate axes of the X-axis, Y-axis, and Z-axis. In Figure 1A, the Z-axis is an axis along the longitudinal direction of the pivot insulator 104, the Y-axis is an axis along the longitudinal direction of the pivot pin 106 which is the pivot axis of the blade 107, and the X-axis is an axis perpendicular to the Z-axis and Y-axis and perpendicular to the longitudinal direction of the pivot pin 106. 1B and 1C, the Z axis is an axis along the longitudinal direction of the fixed insulators 101A and 101B, the Y axis is an axis along the direction in which the fixed insulators 101A and 101B are arranged adjacent to each other, and the X axis is an axis perpendicular to the Z axis and the Y axis and perpendicular to the longitudinal direction of the fixed insulators 101A and 101B. The directions of the X axis, Y axis, and Z axis in Fig. 1A are the same as the directions of the X axis, Y axis, and Z axis in Fig. 1C when the movable contacts 109A and 109B contact the fixed contacts 102A and 102B, respectively.

[0019] The disconnectors 100 and 500 are switches which are set to an open state when disconnecting the electric circuit of a power line when a fault current or the like has been interrupted by the circuit breaker and no load current is flowing, and which are set to a closed state when reconnecting the electric circuit of the disconnected power line.

[0020] The disconnectors 100 and 500 include fixed insulators 101A and 101B, fixed contacts 102A and 102B, pushing pieces 103A and 103B, a rotating insulator 104, a twisting mechanism 105, a rotating pin 106, a blade 107, knock pins 108A and 108B, and movable contacts 109A and 109B.

[0021] The fixed insulators 101A and 101B have a long cylindrical shape and are supported on a support base (not shown) so as to stand vertically along the Z axis. The fixed insulators 101A and 101B are also spaced apart in the direction along the Y axis by a distance that allows the movable contacts 109A and 109B to come into contact with the fixed contacts 102A and 102B, respectively.

[0022] A fixed contactor 102A is coupled to an end of the fixed insulator 101A on the +Z side, which is opposite to the side supported by the support base. The fixed contactor 102A is connected to the power supply side, which is the upstream side of the power line. The fixed contactor 102A includes a metal plate 1021A and a pair of metal plates 1022A and 1023A that guide the movable contactor 109A. The metal plates 1021A, 1022A, and 1023A are formed by bending. The movable contactor 109A comes into contact with the metal plates 1022A and 1023A.

[0023] The metal plate 1021A is, for example, a rectangular flat plate parallel to the YZ plane formed by the Y axis and the Z axis. The metal plate 1022A is, for example, a rectangular flat plate parallel to the XY plane formed by the X axis and the Y axis, extending from one side of the metal plate 1021A parallel to the Y axis on the +Z side toward the +X direction. Here, the tip of the metal plate 1022A on the opposite side to the metal plate 1021A has a shape that bends toward the +Z side as it moves away from the metal plate 1021A to the +X side. The metal plate 1023A is, for example, a rectangular flat plate parallel to the XY plane, extending from one side of the metal plate 1021A parallel to the Y axis on the -Z side toward the +X direction. Here, the tip of the metal plate 1023A on the opposite side to the metal plate 1021A has a shape that bends toward the -Z side as it moves away from the metal plate 1021A to the +X side. In this way, fixed contactor 102A has a U-shape when viewed from the +Y side to the -Y side, and the distance along the Z axis of the tips of a pair of metal plates 1022A, 1023A, which are the tips on the +X side of fixed contactor 102A, is greater than the distance along the Z axis other than the tips of the pair of metal plates 1022A, 1023A, so that movable contactor 109A can be easily guided between the pair of metal plates 1022A, 1023A.

[0024] Similarly, a fixed contactor 102B is coupled to an end of the fixed insulator 101B on the +Z side, which is opposite to the side supported by the support base. The fixed contactor 102B is connected to the load side, which is the downstream side of the power line. The fixed contactor 102B includes a metal plate 1021B and a pair of metal plates 1022B, 1023B that guide the movable contactor 109B. The metal plates 1021B, 1022B, 1023B are formed, for example, by bending a single metal plate. The movable contactor 109B comes into contact with the metal plates 1022B, 1023B.

[0025] The metal plate 1021B is, for example, a rectangular flat plate parallel to the YZ plane. The metal plate 1022B is, for example, a rectangular flat plate parallel to the XY plane, extending from one side of the metal plate 1021B parallel to the Y axis on the -Z side toward the -X direction. Here, the tip of the metal plate 1022B on the opposite side to the metal plate 1021B has a shape that bends toward the -Z side as it moves away from the metal plate 1021B to the -X side. The metal plate 1023B is, for example, a rectangular flat plate parallel to the XY plane, extending from one side of the metal plate 1021B parallel to the Y axis on the +Z side toward the -X direction. Here, the tip of the metal plate 1023B on the opposite side to the metal plate 1021B has a shape that bends toward the +Z side as it moves away from the metal plate 1021B to the -X side. In this way, fixed contactor 102B has a U-shape when viewed from the -Y side toward the +Y side, and the distance along the Z axis of the tips of a pair of metal plates 1022B, 1023B, which are the -X side tips of fixed contactor 102B, is greater than the distance along the Z axis other than the tips of the pair of metal plates 1022B, 1023B, so that movable contactor 109B can be easily guided between the pair of metal plates 1022B, 1023B.

[0026] The fixed contacts 102A and 102B are formed using a metal material such as copper, while the metal plates 1021A, 1022A, and 1023A are formed by silver plating a plate member of a metal such as copper.

[0027] A long-shaped push piece 103A is connected adjacent to the fixed contact 102A. One end of the push piece 103A is connected to the side on the +Y side of the metal plate 1021A, and the push piece 103A extends in the +Z direction along the Z axis, then bends at a right angle at a position on the +Z side of the metal plate 1022A, and extends along the X axis until it is at a position on the +X side of the tips of the metal plates 1022A and 1023A.

[0028] Similarly, a long-shaped push piece 103B is connected adjacent to the fixed contact 102B. One end of the push piece 103B is connected to the -Y side of the metal plate 1021B, and the push piece 103B extends in the -Z direction along the Z axis, then bends at a right angle at a position on the -Z side of the metal plate 1022B, and extends along the X axis until it is at a position on the -X side of the tips of the metal plates 1022B and 1023B.

[0029] The rotating insulator 104 has a long cylindrical shape and is erected vertically along the Z axis at a position midway between the fixed insulators 101A and 101B in the direction along the Y axis, and is supported on a support base so as to rotate around the longitudinal axis of the rotating insulator 104 (axis along the Z axis).

[0030] A twisting mechanism 105 is coupled to the end of the rotating insulator 104 opposite to the support base. A blade 107 having a long, flat shape is coupled to the end of the twisting mechanism 105 opposite to the rotating insulator 104 via a pivot pin 106 arranged parallel to the longitudinal direction of the blade 107 so as to rotate about the longitudinal axis of the pivot pin 106.

[0031] A movable contact 109A that comes into contact with metal plates 1022A and 1023A of fixed contact 102A is coupled to the end of blade 107 on the fixed contact 102A side. Movable contact 109A is, for example, a rectangular metal plate parallel to the flat surface of blade 107. Furthermore, a knock pin 108A that stands in a direction perpendicular to the flat surface of blade 107 and the flat surface of movable contact 109A is provided at the end of blade 107 on the fixed contact 102A side, facing the end of pushing piece 103A.

[0032] Similarly, a movable contact 109B that comes into contact with metal plates 1022B and 1023B of fixed contact 102B is coupled to the end of blade 107 on the fixed contact 102B side. Movable contact 109B is, for example, a rectangular metal plate parallel to the flat surface of blade 107. Furthermore, at the end of blade 107 on the fixed contact 102B side, facing the end of pushing piece 103B, a knock pin 108B is provided that stands in a direction perpendicular to the flat surface of blade 107 and the flat surface of movable contact 109B and in a direction opposite to the direction in which knock pin 108A stands.

[0033] The length from the center of rotation of the blade 107 to the movable contact 109A is equal to the length from the center of rotation of the blade 107 to the movable contact 109B. When the rotating pin 106 is viewed from the movable contact 109B side toward the movable contact 109A side, the blade 107 is elastically biased in the clockwise direction (-R2 direction) around the rotating pin 106. For example, when the knock pins 108A and 108B are separated from the pushing pieces 103A and 103B, respectively, as shown in FIG. 1B, the flat surface of the blade 107 and the flat surfaces of the movable contacts 109A and 109B are parallel to the XY plane due to the elastic force applied to the blade 107, and the knock pin 108A is raised toward the +Z direction and the knock pin 108B is raised toward the -Z direction. At this time, the power line is disconnected. 1B, when the rotating insulator 104 rotates in the R1 direction around the longitudinal axis of the rotating insulator 104, the knock pins 108A and 108B continue to be pressed by the pushing pieces 103A and 103B, respectively, and the blade 107 rotates 90 degrees in the R2 direction around the rotating pin 106 against the elastic force by the operation of the twisting mechanism 105. Then, as shown in FIG. 1C, the flat surface of the blade 107 and the flat surfaces of the movable contacts 109A and 109B become parallel to the YZ plane, the movable contacts 109A and 109B come into contact with the metal plates 1022A and 1023A and the metal plates 1022B and 1023B, respectively, and the knock pin 108A rises in the +X direction and the knock pin 108B rises in the -X direction. At this time, the power line is connected.

[0034] The movable contacts 109A and 109B are formed, for example, by silver plating a plate member made of a metal material such as copper.

[0035] After being installed in the power system, the disconnector 500 is affected by the surrounding environment of the installation location. As a result, as the disconnector 500 ages after installation, dust adhesion, rusting, aging deterioration, etc. occur in the fixed contacts 102A and 102B, the movable contacts 109A and 109B, the blade 107, the twisting mechanism 105 for rotating the blade 107, etc. For example, in the structure of the disconnector 500, the rotating pin 106 connecting the blade 107 and the twisting mechanism 105 is configured to rotate while sliding on the connecting portion of the blade 107 and the twisting mechanism 105 when the blade 107 rotates in either the R2 direction or the -R2 direction so that the blade 107 can rotate in the R2 direction and the -R2 direction. Therefore, if dust adheres to at least one of the blade 107, the pivot pin 106, and the twisting mechanism 105, rust occurs, or deterioration over time such as galling or discoloration occurs, which indicates metal fatigue, the blade 107 may malfunction and become difficult to rotate between the positions of FIG. 1B and FIG. 1C. If the disconnector 500 malfunctions in this way, it may not be possible to normally disconnect or connect the power line. In particular, if it is found that the disconnector 500 has already malfunctioned at the time when the blade 107 is rotated from the state of FIG. 1B to the state of FIG. 1C to open the disconnector 500, the disconnector 500 must be inspected and repaired as necessary from this point in time, and it takes a lot of time to open the disconnector 500. This problem cannot be solved by regular inspection of the disconnector 500 alone.

[0036] In order to prevent such problems from occurring, the prediction system 1 of this embodiment predicts the probability of the disconnector 500 malfunctioning at the scheduled time of operation using AI (Artificial Intelligence) prior to the scheduled time of operation of the disconnector 500 (for example, 3 months, 6 months, 1 year, etc. prior to the scheduled time of operation), and is a system that can prompt inspection of the disconnector 500 before the scheduled time of operation of the disconnector 500, for example, if the numerical value of the probability of the disconnector 500 malfunctioning is greater than the standard numerical value of the probability that inspection or repair of the disconnector 100 is required.

[0037] FIG. 2 is a block diagram showing a schematic configuration of a prediction system 1 including an information processing device 200 that predicts the probability that the disconnector 500 will malfunction at the scheduled operation time by using a learning model (AI model).

[0038] The prediction system 1 includes an information processing device 200 and an external computer 300 as means for realizing the above functions. The information processing device 200 and the external computer 300 are connected in a state where two-way communication is possible via a communication network 400. The communication network 400 is, for example, a local area network (LAN), a wide area network (WAN), a dedicated line, a power line communication network, various public communication networks, etc.

[0039] The information processing device 200 includes an input unit 210 , a prediction unit 220 , an output unit 230 , and a storage unit 240 .

[0040] The memory unit 240 stores a control program 241, learning data 242, and a learning model 243.

[0041] The control program 241 is a program executed by the information processing device 200 when causing the prediction unit 220 to predict the probability that the disconnector 500 will malfunction at the scheduled operation time, prior to the scheduled operation time of the disconnector 500. Note that the scheduled operation time of the disconnector 500 may be, for example, the time of the next regular inspection to check whether the disconnector 500 operates normally, or may be the time to set the disconnector 500 to an open state in accordance with the scheduled time for replacement work of the power line.

[0042] The learning data 242 is supervised data in which feature quantities and labels, which are past performance data of the disconnector 100, are associated with each other, and is prepared in advance to enable the prediction unit 220 to predict the probability that the current disconnector 500 will malfunction at the scheduled operation time. The learning data 242 is data prepared in advance, for example, for a plurality of past disconnectors 100 of the same type (specifications) as the current disconnector 500 and manufactured in the same manufacturing year, so that the learning model 243 can predict the probability that the disconnector 500 will malfunction at the scheduled operation time.

[0043] FIG. 3 is a diagram showing an example of features and labels constituting the training data 242.

[0044] The feature amount includes data indicating, for example, (A) images taken at predetermined intervals (e.g., once a year) since the disconnector 100 was installed in the power system in the past, (B) the installation location of the disconnector 100, and (C) the number of times the disconnector 100 operated at each predetermined interval in the past since the disconnector 100 was installed in the power system. In addition, the label includes (D) data indicating the period during which the disconnector 100 was operable in the past from its installation in the power system until it malfunctioned.

[0045] (A) The images taken at predetermined intervals in the past since the disconnector 100 was installed in the power system are images taken of the blade 107, the twisting mechanism 105, and the pivot pin 106 from one direction of a predetermined white arrow shown in Fig. 1B so that, for example, the adhesion of dust, rust, deterioration over time, etc. to the blade 107, the twisting mechanism 105, and the pivot pin 106, which are components of the disconnector 100, can be simultaneously seen. Data showing this image for one disconnector 100 is continuously taken at predetermined intervals until the disconnector 100 malfunctions, and is accumulated as a feature amount. Since the direction in which the disconnector 100 is photographed is determined to be one and the same direction, the learning model 243 can accurately learn the relationship between the state of change such as adhesion of dust, rust, and aging deterioration of the blade 107, the twisting mechanism 105, and the rotating pin 106 for each predetermined period, and the operable period from the installation of the disconnector 100 until the disconnector 100 malfunctions. Note that the direction in which the disconnector 100 is photographed is preferably the same for any predetermined period, but may be slightly deviated from the direction of the white arrow as long as the blade 107, the twisting mechanism 105, and the rotating pin 106 can be included in one image. In addition, the image may be an image photographed so that the fixed contacts 102A, 102B, and the movable contacts 109A, 109B are included in the image together with the blade 107, the twisting mechanism 105, and the rotating pin 106.

[0046] 4 shows an example of an image actually taken of the blade 107, twisting mechanism 105, and pivot pin 106 of the disconnector 100. From this image, it can be seen that the learning model 243 is capable of learning the state of changes such as adhesion of dust, rust, and aging deterioration of the blade 107, twisting mechanism 105, and pivot pin 106.

[0047] (B) The installation location of the disconnector 100 is included in the feature in the form of data indicating, for example, the latitude and longitude of the location where the disconnector 100 is installed. The disconnector 100 is affected by various factors depending on the surrounding environment of the installation location. For example, when the disconnector 100 is installed in a power system in a coastal area, the disconnector 100 is more susceptible to damage such as salt damage and rust than a disconnector 100 installed in a power system in a non-coast area, and the operational period from installation to malfunction tends to be shorter. Therefore, by including data indicating the installation location of the disconnector 100 in the feature, the learning model 243 can accurately learn the relationship between the state of dust adhesion, rust, aging deterioration, etc. of the blade 107, the twisting mechanism 105, and the pivot pin 106 for each predetermined period and the operational period from the installation of the disconnector 100 to the malfunction of the disconnector 100.

[0048] (C) The number of operations per predetermined period after the disconnector 100 is installed in the power system is included in the feature as data indicating, for example, how many times the disconnector 100 performs an operation in which a combination of an opening operation and a closing operation following the opening operation is counted as one operation per predetermined period. The disconnector 100 has a tendency that the operable period from installation to malfunction becomes shorter as the number of opening and closing operations increases. Therefore, by including data indicating the number of operations per predetermined period after the disconnector 100 is installed in the power system in the feature, the learning model 243 can accurately learn the relationship between the state of dust adhesion, rusting, aging deterioration, etc. of the blade 107, the twisting mechanism 105, and the pivot pin 106 per predetermined period and the operable period from the installation of the disconnector 100 in the power system to the disconnector 100 malfunctioning.

[0049] In this embodiment, the feature includes both data indicating (B) the installation location of the disconnector 100 and (C) the number of times the disconnector 100 operates per specified period since it was installed in the power grid, but it may be configured such that only one of the data is included in the feature.

[0050] (D) The operable period from when the disconnector 100 is installed in the power grid until when it malfunctions is included in the label as data associated with the features (A), (B), and (C).

[0051] The learning model 243 is a machine learning model formed to predict the probability that the disconnector 500 will malfunction at the scheduled operation time by learning using the learning data 242, when data indicating an image of the blade 107, the twisting mechanism 105, and the pivot pin 106 in the disconnector 500 taken from the direction of the white arrow in FIG. 1B, the time elapsed since the disconnector 500 was installed in the power system, the scheduled operation time of the disconnector 500, the installation location of the disconnector 500, and the number of operations of the disconnector 500 up to now are input. The more learning data 242 the learning model 243 uses to learn, the more accurate the prediction becomes. In this embodiment, the learning model 243 is, for example, a deep neural network (DNN), but may be another type of model such as gradient boosting (GBDT).

[0052] FIG. 5 is a diagram showing the structure of a deep neural network, which is an example of the learning model 243.

[0053] The learning model 243 includes three layers: an input layer 2431, an intermediate layer 2432, and an output layer 2433. The input layer 2431 is of the same type as the feature amounts (A), (B), and (C) constituting the learning data 242, and is a layer to which data indicating images of the blade 107, the twisting mechanism 105, and the pivot pin 106 of the disconnector 500, the period of time since the disconnector 500 was installed in the power system, the planned operation time of the disconnector 500, the installation location of the disconnector 500, and the number of operations of the disconnector 500 up to now are input. The intermediate layer 2432 is a layer including one or more hidden layers consisting of one or more nodes including parameters adjusted by performing learning using the learning data 242. The intermediate layer 2432 predicts the probability that the disconnector 500 will malfunction at the planned operation time based on the data input to the input layer 2431. The output layer 2433 is a layer that outputs the prediction result by the intermediate layer 2432.

[0054] The learning model 243 refers to the learning data 242 of a similar disconnector 100 for all data, such as images of the blade 107, twisting mechanism 105, and pivot pin 106 of the disconnector 500, the period since the disconnector 500 was installed in the power system, the installation location of the disconnector 500, and the number of operations of the disconnector 500 up to now, and predicts the probability that the disconnector 500 will malfunction at the scheduled operation time. Note that the installation location of the disconnector 100 referred to for the installation location of the disconnector 500 may include, in addition to the surrounding locations including the installation location of the disconnector 500, locations that have the same or similar environment as the surrounding environment including the installation location of the disconnector 500.

[0055] 2, when the information processing device 200 predicts the probability that the disconnector 500 will malfunction at the scheduled operation time, the input unit 210 receives as input images of the blade 107, twisting mechanism 105, and pivot pin 106 of the disconnector 500, the period of time that has elapsed since the disconnector 500 was installed in the power grid, the installation location of the disconnector 500, and data indicating the number of operations of the disconnector 500 to date. The data input to the input unit 210 is then input to the prediction unit 220.

[0056] The prediction unit 220 inputs the above data input to the input unit 210 into the input layer 2431 of the learning model 243, and instructs the learning model 243 to predict the probability that the disconnector 500 will malfunction at the scheduled operation time. Here, the storage unit 240 stores a reference probability value for determining whether the probability that the disconnector 500 is predicted to malfunction at the scheduled operation time is higher than the probability that the disconnector 100 needs to be repaired. When the learning model 243 predicts the probability that the disconnector 500 will malfunction at the scheduled operation time, the prediction unit 220 compares the value of the probability that the disconnector 500 will malfunction at the scheduled operation time with the reference probability value read from the storage unit 240, and creates a message that prompts the disconnector 500 to be inspected if the value of the probability that the disconnector 500 will malfunction at the scheduled operation time is greater than the reference probability value.

[0057] The output unit 230 outputs the prediction result by the learning model 243 to the outside in accordance with an instruction from the prediction unit 220 triggered by the output of the prediction result from the intermediate layer 2432 to the output layer 2433. In addition, when the numerical value of the probability that the disconnector 500 will malfunction at the scheduled operation time is greater than the numerical value of the reference probability, the output unit 230 also outputs a message created by the prediction unit 220 to encourage inspection of the disconnector 500. Possible output methods of the output unit 230 include, for example, a method of displaying and outputting the prediction result and the message on a display panel, a method of outputting the prediction result and the message as voice from a speaker, a method of printing and outputting the prediction result and the message on a paper medium, and the like.

[0058] The external computer 300 has a memory unit 310, and stores in the memory unit 310 data indicating the probability that the circuit breaker 500 will malfunction at the scheduled operation time predicted by the learning model 243, and a message created by the prediction unit 220 to encourage inspection of the circuit breaker 500, and reads out these data from the memory unit 310 and outputs them to the information processing device 200 as necessary.

[0059] FIG. 6 is a block diagram showing an example of hardware of an information processing device 200 used to realize the prediction system 1. As shown in FIG.

[0060] The information processing device 200 includes a processor 2010, a main memory device 2020, an auxiliary memory device 2030, an input device 2040, an output device 2050, and a communication device 2060. The information processing device 200 is, for example, a personal computer, an office computer, various server devices, a general-purpose machine, or the like. The information processing device 200 may be realized, in whole or in part, by using virtual information processing resources provided by using a virtualization technology, such as a virtual server provided by a cloud system. The prediction system 1 may be realized by using a plurality of information processing devices 200 connected to each other so as to be able to communicate with each other.

[0061] The processor 2010 is configured using, for example, a Central Processing Unit (CPU), a Micro Processing Unit (MPU), a Graphics Processing Unit (GPU), a Field Programmable Gate Array (FPGA), an Application Specific Integrated Circuit (ASIC), an Artificial Intelligence (AI) chip, etc.

[0062] The main memory device 2020 is a device that stores programs and data, and is, for example, a Read Only Memory (ROM), a Random Access Memory (RAM), or a non-volatile memory (NVRAM (Non Volatile RAM)).

[0063] The auxiliary storage device 2030 is, for example, a solid state drive (SSD), a hard disk drive, an optical storage device (CD (Compact Disc), DVD (Digital Versatile Disc), etc.), a storage system, an IC card, an SD card, a read / write device for a recording medium such as an optical recording medium, a storage area of ​​a cloud server, etc. Programs and data can be read into the auxiliary storage device 2030 via a recording medium reader or a communication device 2060. The programs and data stored in the auxiliary storage device 2030 are read into the main storage device 2020 as needed.

[0064] The input device 2040 is an interface that accepts input from the outside, and is, for example, a keyboard, a mouse, a touch panel, a card reader, a pen-input tablet, a voice input device, or the like.

[0065] The output device 2050 is an interface that outputs various information such as the progress of processing and processing results. The output device 2050 is, for example, a display device (LCD (Liquid Crystal Display), graphic card, etc.) that visualizes the various information described above, a device that converts the various information described above into voice (voice output device (speaker, etc.)), or a device that converts the various information described above into text (printer, etc.). The information processing device 200 may be configured to input and output information to and from other devices via the communication device 2060.

[0066] The input device 2040 and the output device 2050 constitute a user interface that receives and presents information to and from the user.

[0067] The communication device 2060 is a device that realizes communication (wired communication or wireless communication) with other devices via a communication infrastructure such as a communication network 400, and is configured using, for example, a NIC (Network Interface Card), a wireless communication module, a USB module, etc.

[0068] In addition, the information processing device 200 may be implemented with, for example, an operating system, a file system, a DBMS (Data Base Management System) (relational database, NoSQL, etc.), a KVS (Key-Value Store), etc.

[0069] The functions of the prediction system 1 are realized by the processor 2010 of the information processing device 200 executing a program (control program) loaded into the main memory device 2020, or by the functions of the hardware (FPGA, ASIC, AI chip, etc.) constituting the prediction system 1. For example, the memory unit 240 in FIG. 2 is realized by the main memory device 2020 and the auxiliary memory device 2030, the input unit 210 in FIG. 2 is realized by the input device 2040, the prediction unit 220 in FIG. 2 is realized by the processor 2010, and the output unit 330 in FIG. 2 is realized by the output device 2050 and the communication device 2060.

[0070] FIG. 7 is a flowchart showing a process performed by the information processing device 200 to learn the learning model 243 using the learning data 242.

[0071] First, the information processing device 200 generates learning data 242 that corresponds features including data indicating (A) images taken at predetermined intervals since the disconnector 100 was installed in the power system, (B) the installation location of the disconnector 100, and (C) the number of times the disconnector 100 operates at each predetermined interval since the disconnector 100 was installed in the power system, and (D) labels including data indicating the period of time during which the disconnector 100 can operate from its installation in the power system until it malfunctions (S1000).

[0072] Next, the information processing device 200 uses the learning data 242 to train the learning model 243 (S1010). Note that the information processing device 200 may verify the prediction accuracy of the trained learning model 243. In this case, the information processing device 200 prepares learning data and verification data as the learning data 242, trains the learning model 243 using the learning data, and verifies the learning model 243 using the verification data.

[0073] FIG. 8 is a flowchart showing a process in which the information processing device 200 uses the learning model 243 to predict the probability that the disconnector 500 will malfunction at the scheduled operation time.

[0074] First, the information processing device 200 judges whether or not all data regarding the disconnector 500, such as images of the blade 107, the twisting mechanism 105, and the pivot pin 106, the time since the disconnector 500 was installed in the power grid, the installation location, and the number of operations up to now, have been input to the input unit 210 (S2000). When the information processing device 200 judges that the above data has not been input to the input unit 210 (S2000: NO), it executes the process of step S2000 again.

[0075] Next, when the information processing device 200 determines that the above data has been input to the input unit 210 (S2000: YES), it predicts the probability that the disconnector 500 will malfunction at the scheduled operation time using the learning model 243 (S2010).

[0076] Next, the information processing device 200 determines whether the numerical value of the probability that the disconnector 500 will malfunction at the scheduled operation time is greater than the numerical value of the reference probability (S2020).

[0077] If the numerical value of the probability that the disconnector 500 will malfunction at the scheduled operation time is greater than the numerical value of the reference probability (S2020: YES), the information processing device 200 outputs the predicted result of the probability that the disconnector 500 will malfunction at the scheduled operation time and a message prompting the user to inspect the disconnector 500 (S2030).

[0078] On the other hand, if the numerical value of the probability that the circuit breaker 500 will malfunction at the scheduled operation time is smaller than the numerical value of the reference probability (S2020: NO), the information processing device 200 outputs only the predicted result of the probability that the circuit breaker 500 will malfunction at the scheduled operation time (S2040).

[0079] As described above, the information processing device 200 having the processor 2010 and the memory device (main memory device 2020, auxiliary memory device 2030) stores a learning model 243 that performs learning based on learning data 242 in which feature quantities including data of images of the blade 107, twisting mechanism 105, and pivot pin 106 of the disconnector 100 taken in the past at predetermined intervals since the installation of the disconnector 100 are associated with labels including data indicating the operable period from the installation of the disconnector 100 until it malfunctions, and currently inputs data of images of the blade 107, twisting mechanism 105, and pivot pin 106 of the disconnector 500 taken after the installation of the disconnector 500, the period elapsed since the installation of the disconnector 500, and the planned operation time of the disconnector 500 into the learning model 243, thereby predicting the probability that the disconnector 500 will malfunction at the planned operation time.

[0080] According to the information processing device 200, the predicted result of the probability that the disconnector 500 will malfunction at the scheduled operation time can be confirmed before the scheduled operation time of the disconnector 500, so that inspection work can be performed on the disconnector 500 according to the probability that the disconnector 500 will malfunction.

[0081] Furthermore, in the information processing device 200, the feature quantity includes data indicating the installation location of the disconnector 100, and currently, by inputting an image of the disconnector 500, data indicating the period of time that has elapsed since the disconnector 500 was installed in the power system, the planned time to operate the disconnector 500, and the installation location of the disconnector 500 into the learning model 243, the probability that the disconnector 500 will malfunction at the planned time of operation is predicted.

[0082] In general, the causes of the disconnector 500 malfunctioning vary depending on the environment at the installation location of the disconnector 500, and the operable period from when the disconnector 500 is installed in the power system until it malfunctions also varies depending on the environment at the installation location of the disconnector 500. According to the information processing device 200, the learning model 243 learns using a combination of features and labels including data indicating the installation location of the disconnector 100, and data indicating the installation location of the disconnector 500 is also input to the learning model 243, making it possible to improve the prediction accuracy when predicting the probability that the disconnector 500 will malfunction at the scheduled operation time.

[0083] Furthermore, in the information processing device 200, the feature quantity includes data indicating the number of times the disconnector 100 has operated for each specified period since it was installed in the power system, and by inputting data indicating the current image of the disconnector 500, the period elapsed since the disconnector 500 was installed in the power system, the planned time for operating the disconnector 500, the installation location of the disconnector 500, and the number of times the disconnector 500 has operated up to now into the learning model 243, the probability that the disconnector 500 will malfunction at the planned time for operation is predicted.

[0084] The operable period from when the disconnector 500 is installed in the power system until it malfunctions tends to be shorter as the number of operations of the disconnector 500 since it is installed in the power system increases. According to the information processing device 200, the learning model 243 learns by combining features and labels including data indicating the number of operations for each predetermined period since the disconnector 100 is installed in the power system, and data indicating the number of operations from when the disconnector 500 is installed in the power system to the present is also input to the learning model 243, so that it is possible to improve the prediction accuracy when predicting the probability that the disconnector 500 will malfunction at the scheduled operation time.

[0085] In addition, in the information processing device 200, the disconnectors 100, 500 are the same type of disconnectors, and the images of the disconnectors 100, 500 are images taken from the same direction with respect to the disconnectors 100, 500, respectively.

[0086] According to the information processing device 200, the dust adhesion, rust, deterioration over time, etc. occurring in the current disconnector 500 can be easily compared with the dust adhesion, rust, deterioration over time, etc. occurring in the past disconnector 100, which makes it possible to improve the prediction accuracy when predicting the probability that the disconnector 500 will malfunction at the scheduled operation time.

[0087] The above-described embodiment is for the purpose of facilitating understanding of the present invention, and is not intended to limit the present invention. The present invention may be modified or improved without departing from the spirit of the present invention, and equivalents thereof are also included in the present invention. [Explanation of symbols]

[0088] 1. Prediction System 100 Disconnector 101A, 101B Fixed insulator 102A,102B Fixed contact 103A,103B Push piece 104 Rotating insulator 105 Twisting mechanism 106 Pivot pin 107 Blade 108A, 108B Knock pin 109A,109B Movable contact 200 Information processing device 210 Input section 220 Prediction Department 230 Output section 240,310 storage section 241 Control Program 242 Training Data 243 Learning Model 300 External Computer 400 Communication Network 1021A,1022A,1023A,1021B,1022B,1023B Metal plate 2010 Processors 2020 Main storage 2030 Auxiliary storage 2040 Input Device 2050 Output Device 2060 Communication Equipment

Claims

1. A processor and a storage device are included. storing a learning model that performs learning based on learning data in which a feature amount including data showing a first image of the first disconnecting switch taken at predetermined intervals in the past since the first disconnecting switch was installed is associated with a label including data showing an operable period from the installation of the first disconnecting switch until it malfunctions; At present, a probability that the second disconnector will malfunction at the scheduled operation time is predicted by inputting data indicating a second image of the second disconnector taken after the second disconnector is installed, a period of time elapsed since the second disconnector is installed, and a scheduled operation time of the second disconnector into the learning model. Information processing device.

2. 2. The information processing device according to claim 1, The feature amount includes data indicating an installation location of the first disconnector, At present, a probability that the second disconnector will malfunction at the scheduled operation time is predicted by inputting data indicating a second image of the second disconnector, a period of time elapsed since the second disconnector was installed, a scheduled operation time of the second disconnector, and an installation location of the second disconnector into the learning model. Information processing device.

3. 2. The information processing device according to claim 1, The characteristic amount includes data indicating the number of operations of the first disconnector per predetermined period, At present, a probability that the second disconnector will malfunction at the scheduled operation time is predicted by inputting data indicating a second image of the second disconnector, a period of time elapsed since the second disconnector was installed, a scheduled operation time of the second disconnector, and a current number of operations of the second disconnector into the learning model. Information processing device.

4. 2. The information processing device according to claim 1, The first disconnector and the second disconnector are the same type of disconnector, The first image and the second image are images of the first disconnector and the second disconnector taken from the same direction, respectively. Information processing device.

5. An information processing method of an information processing device having a processor and a storage device, storing a learning model that performs learning based on learning data in which a feature amount including data showing a first image of the first disconnecting switch taken at predetermined intervals in the past since the first disconnecting switch was installed is associated with a label including data showing an operable period from the installation of the first disconnecting switch until it malfunctions; At present, a second image of the second disconnector taken after the second disconnector is installed, a period of time elapsed since the second disconnector is installed, and data indicating a planned time for operating the second disconnector are input to the learning model to predict the probability that the second disconnector will malfunction at the planned time for operation. Information processing methods.

6. 6. The information processing method according to claim 5, The feature amount includes data indicating an installation location of the first disconnector, At present, a probability that the second disconnector will malfunction at the scheduled operation time is predicted by inputting data indicating a second image of the second disconnector, a period of time elapsed since the second disconnector was installed, a scheduled operation time of the second disconnector, and an installation location of the second disconnector into the learning model. Information processing methods.

7. 6. The information processing method according to claim 5, The characteristic amount includes data indicating the number of operations of the first disconnecting switch per predetermined period, At present, a probability that the second disconnector will malfunction at the scheduled operation time is predicted by inputting data indicating a second image of the second disconnector, a period of time elapsed since the second disconnector was installed, a scheduled operation time of the second disconnector, and a current number of operations of the second disconnector into the learning model. Information processing methods.

8. 6. The information processing method according to claim 5, The first disconnector and the second disconnector are the same type of disconnector, The first image and the second image are images of the first disconnector and the second disconnector taken from the same direction, respectively. Information processing methods.

9. A computer having a processor and a storage device, a first process for storing a learning model that performs learning based on learning data in which a feature amount including data indicating a first image of the first disconnecting switch taken at predetermined intervals in the past since the first disconnecting switch was installed and a label including data indicating an operable period from the installation of the first disconnecting switch until it malfunctions are associated with each other; a second process of predicting a probability that the second disconnector will malfunction at the planned operation time by inputting data indicating a second image of the second disconnector taken after the second disconnector is currently installed, a period of time elapsed since the second disconnector is installed, and a planned operation time of the second disconnector into the learning model; A program that executes the following.

10. The program according to claim 9, The feature amount includes data indicating an installation location of the first disconnector, In the second process, At present, a probability that the second disconnector will malfunction at the scheduled operation time is predicted by inputting data indicating a second image of the second disconnector, a period of time elapsed since the second disconnector was installed, a scheduled operation time of the second disconnector, and an installation location of the second disconnector into the learning model. program.

11. The program according to claim 9, The characteristic amount includes data indicating the number of operations of the first disconnector per predetermined period, In the second process, At present, a probability that the second disconnector will malfunction at the scheduled operation time is predicted by inputting data indicating a second image of the second disconnector, a period of time elapsed since the second disconnector was installed, a scheduled operation time of the second disconnector, and a current number of operations of the second disconnector into the learning model. program.

12. The program according to claim 9, The first disconnector and the second disconnector are the same type of disconnector, The first image and the second image are images of the first disconnector and the second disconnector taken from the same direction, respectively. program.

Images