Machine learning device and temperature anomaly determination device

Abstract

To provide a machine learning device capable of constructing a learning model for more efficiently determining a temperature anomaly at a connection part between a monitored object and an electric wire, and a temperature anomaly determination device using the learning model.SOLUTION: A machine learning device 10 constructs a learning model for determining a temperature anomaly at a connection part 5 between a monitored object 3 and an electric wire 4, and includes: an input data acquisition unit 11 that acquires, as input data, a thermal image including the connection part 5 as a subject; a label acquisition unit 12 that acquires a level of the temperature anomaly of the connection part 5 as a label; and a learning model construction unit 14 that constructs a learning model for determining the level of the temperature anomaly of the connection part 5 included as the subject in the thermal image by performing supervised learning using a pair of the input data and the label as teacher data.SELECTED DRAWING: Figure 3

Description

[Technical field]

[0001] The present invention relates to a machine learning device and a temperature anomaly determination device. [Background technology]

[0002] Conventionally, a technology is known in which a thermal image of a monitored object connected to an electric wire is acquired to judge a temperature abnormality. For example, Patent Documents 1 and 2 describe this type of technology. Patent Document 1 describes a device that estimates the presence or absence of a temperature abnormality and its cause based on a captured thermal image and a temperature distribution pattern stored in a database. Patent Document 2 describes a device that diagnoses a temperature abnormality based on a captured thermal image, the allowable temperature of the object to be diagnosed, and the ambient temperature and humidity. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2012-32405 A [Patent Document 2] JP 2008-45888 A Summary of the Invention [Problem to be solved by the invention]

[0004] Meanwhile, in electric power facilities such as power plants and substations, in order to check the degree of deterioration of the connection between the equipment and electric wires as the monitored object and the presence or absence of connection abnormality, thermal images of the connection are acquired and temperature abnormality of the connection is determined from the thermal images. Since there are many connections between the monitored object and the electric wires in electric power facilities, it takes time and effort to analyze thermal images of multiple connections to determine the presence or absence of abnormality. According to Patent Document 1, it is possible to determine temperature abnormality in a thermal image by taking into account the temperature distribution pattern, and according to Patent Document 2, it is possible to determine temperature abnormality in a thermal image by taking into account the ambient temperature, etc., but there is room for improvement in terms of more efficiently determining temperature abnormality of the connection.

[0005] The present invention aims to provide a machine learning device that can construct a learning model for more efficiently determining temperature anomalies at the connection between an object to be monitored and an electric wire, and a temperature anomaly determination device that uses the learning model. [Means for solving the problem]

[0006] The present invention relates to a machine learning device that constructs a learning model for determining temperature abnormalities at a connection between a monitored object and an electric wire, the machine learning device including an input data acquisition unit that acquires a thermal image including the connection as a subject as input data, a label acquisition unit that acquires a temperature abnormality level at the connection as a label, and a learning model construction unit that constructs a learning model for determining the temperature abnormality level at the connection included as a subject in the thermal image by performing supervised learning using a pair of the input data and the label as teacher data.

[0007] The input data acquisition unit further acquires, as the input data, a current value flowing through the connection portion and meteorological information.

[0008] The object to be monitored is connected to a three-phase AC electric wire, and the thermal image acquired by the input data acquisition unit includes, as subjects, the connection parts between the object to be monitored and all three phases of the electric wires.

[0009] The present invention also relates to a temperature anomaly determination device that uses the learning model constructed by the machine learning device, and includes a thermal image acquisition unit that acquires a new thermal image including a connection portion between the monitored object and the electric wire as a subject, and a temperature anomaly determination unit that determines the level of temperature anomaly at the connection portion between the monitored object and the electric wire based on the thermal image and the learning model. Effect of the Invention

[0010] According to the present invention, it is possible to provide a machine learning device capable of constructing a learning model for more efficiently determining temperature anomalies at the connection between a monitored object and an electric wire, and a temperature anomaly determination device using the learning model. [Brief description of the drawings]

[0011] [Figure 1] 1 is a perspective view showing a monitored object of an electric power facility in which a temperature anomaly determination system according to an embodiment of the present invention is used; [Diagram 2] 1 is a diagram showing an overall configuration of a temperature anomaly determination system according to an embodiment of the present invention; [Diagram 3] FIG. 1 is a functional block diagram of a machine learning device according to an embodiment of the present invention. [Figure 4] FIG. 2 is a diagram showing input data and labels of a machine learning device according to an embodiment of the present invention. [Diagram 5] 1 is a diagram showing a thermal image of a monitored object of an electric power facility in which a temperature anomaly determination system according to an embodiment of the present invention is used. [Figure 6] 1 is a functional block diagram of a temperature anomaly determination device according to an embodiment of the present invention; [Figure 7] FIG. 2 is a flowchart illustrating the operation of a machine learning device according to an embodiment of the present invention. [Figure 8] 4 is a flowchart showing the operation of the temperature anomaly determination device according to the embodiment of the present invention; FIG. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0012] Hereinafter, preferred embodiments of the present invention will be described with reference to the drawings. Fig. 1 is a perspective view showing a monitored object 3 of an electric power facility 2 in which a temperature anomaly determination system 1 according to this embodiment is used. Fig. 2 is a diagram showing the overall configuration of the temperature anomaly determination system 1.

[0013] The temperature anomaly determination system 1 is a system for determining a temperature anomaly at a connection 5 between a monitored object 3 and an electric wire 4 provided in an electric power facility 2. First, the electric power facility 2 in which the temperature anomaly determination system 1 is used will be described.

[0014] The power facility 2 is a facility for supplying three-phase AC power. Examples of the power facility 2 include a power plant and a substation. The power facility 2 is provided with a three-phase AC electric wire 4 and a plurality of monitored objects 3, which are devices connected to the electric wire 4. Examples of the types of monitored objects 3 include a transformer and a switchgear.

[0015] The connection portion 5 between the monitored object 3 and the electric wire 4 includes a plate-shaped equipment side terminal portion 51 provided at an end of the monitored object 3, and a plate-shaped electric wire side terminal portion 52 provided at an end of the electric wire 4. A plurality of terminal holes are provided in each of the equipment side terminal portion 51 and the electric wire side terminal portion 52. The monitored object 3 and the electric wire 4 are connected by overlapping the equipment side terminal portion 51 and the electric wire side terminal portion 52 and fastening them with bolts.

[0016] As shown in FIG. 1, for example, the monitored object 3 has six connection parts 5, ie, connection parts 5a to 5f, which are connection parts 5 with the electric wires 4. Specifically, one monitored object 3 has connection parts 5a, 5b, and 5c with the electric wires 4 of each phase of the three-phase AC at its upstream end in the power transmission direction, and connection parts 5d, 5e, and 5f with the electric wires 4 of each phase of the three-phase AC at its downstream end in the power transmission direction. That is, the monitored object 3 has three connection parts 5 with one set of electric wires 4. In this way, the monitored object 3 has many connection points with the electric wires 4. Since the electric power facility 2 has many monitored objects 3, there are many connection parts 5 to be judged by the temperature anomaly judgment system 1.

[0017] Next, a description will be given of the configuration of the temperature anomaly determination system 1. As shown in Fig. 2, the temperature anomaly determination system 1 includes a machine learning device 10 and a temperature anomaly determination device 20.

[0018] Here, the machine learning device 10 and the temperature anomaly determination device 20 are paired one-to-one and connected to be able to communicate with each other. Although not shown in FIG. 1, the machine learning device 10 and the temperature anomaly determination device 20 may be connected to each other via a network. The network may be, for example, a local area network (LAN), the Internet, a public telephone network, or a combination of these. There are no particular limitations on the specific communication method in the network, whether it is a wired connection or a wireless connection, etc. Alternatively, the machine learning device 10 and the temperature anomaly determination device 20 may be directly connected via a connector, rather than communicating using a network.

[0019] The machine learning device 10 will be described with reference to Figs. 3 to 5. Fig. 3 is a diagram showing functional blocks of the machine learning device 10. Fig. 4 is a diagram showing input data and labels acquired by the machine learning device 10. Fig. 5 is a diagram showing a thermal image of a monitored object 3 of an electric power facility 2 in which a temperature anomaly determination system 1 is used. Note that in Fig. 5, of all three phases of electric wires 4 of a three-phase AC and the connection portion 5 between the electric wires 4 and the monitored object 3 for two phases, illustration of the connection portion 5 between the electric wires 4 and the monitored object 3 is omitted.

[0020] The machine learning device 10 constructs a learning model for use in the temperature anomaly determination device 20 by supervised learning. The learning model constructed by the machine learning device 10 is used to determine a temperature anomaly at the connection portion 5 between the monitored object 3 and the electric wire 4. As shown in FIG. 3 , the machine learning device 10 includes an input data acquisition unit 11, a label acquisition unit 12, a memory unit 13, and a learning model construction unit 14.

[0021] The input data acquiring unit 11 acquires, as input data, various types of information related to the connection unit 5. As shown in Fig. 4, the input data acquired by the input data acquiring unit 11 includes a thermal image, a connection unit identification number, current data, and weather data.

[0022] A thermal image is an image that includes the connection parts 5 as a subject, as shown in Fig. 5, and shows the temperature distribution of the subject. For example, a thermal image may be an image of the connection parts 5 captured by an infrared camera or the like. In this embodiment, the thermal image includes the connection parts 5 between the monitored object 3 and all three-phase electric wires 4 as a subject. That is, one thermal image identifies temperature data of the connection parts 5 for the three phases and temperature difference data between the connection parts 5 for the three phases (hereinafter referred to as "three-phase temperature difference data").

[0023] Here, in the electric power equipment 2, if there occurs a poor connection between the monitored object 3 and the electric wire 4, or deterioration such as peeling of the plating covering the surface of the equipment side terminal portion 51 or the electric wire side terminal portion 52, the connection portion 5 will overheat. Therefore, if the temperature data of the connection portion 5 obtained from the thermal image is high, it can be determined that there is a high possibility that there is a poor connection between the monitored object 3 and the electric wire 4, or deterioration of the connection portion 5. Also, if there is a large temperature difference between the connection portions 5 with each of the three-phase electric wires 4, it can be determined that there is a high possibility that there is a poor connection, deterioration, or the like in the connection portion 5 with any of the three-phase electric wires 4.

[0024] The connection part identification number is an identification number given to the connection part 5 in the power equipment 2. One connection part identification number is given to each of the three connections 5 between the monitored object 3 and all three-phase electric wires 4. For example, in the monitored object 3 shown in Fig. 1, one connection part identification number is assigned to each of the connections 5a to 5c. That is, by linking the connection part identification number with the thermal image, the connection part 5 shown in the thermal image can be identified.

[0025] The current data is a value of a current flowing through the connection portion 5 included in the thermal image. The current data is acquired, for example, from a monitoring system that monitors the current flowing through the power equipment 2.

[0026] The weather data means weather information for the area where the power facility 2 is located. Specifically, the weather data includes temperature data, humidity data, and sunshine data. For example, the temperature data and humidity data may be the temperature and humidity at the time when the thermal image is acquired, or may be the average temperature and average humidity for one day. Also, they may be the temperature and humidity around the connection unit 5. The sunshine data may be the amount of sunshine on the connection unit 5 at the time when the thermal image is acquired, or may be the average amount of sunshine irradiated on the connection unit 5 for one day. These data may be values ​​obtained from sensors or measuring instruments that measure the temperature, humidity, and amount of sunshine, or may be values ​​announced by a meteorological observation station.

[0027] The label acquisition unit 12 acquires the degree (level) of temperature abnormality of the connection portion 5 included as a subject of the thermal image as a label. The level of temperature abnormality of the connection portion 5 may be set by, for example, a skilled worker analyzing the thermal image, meteorological data, etc.

[0028] The level of temperature abnormality may be, for example, a score indicated by a numerical value, or may be a stepwise determination of the degree of temperature abnormality using letters or symbols such as excellent, good, acceptable, or unacceptable. In this embodiment, four levels, Level A to Level D, are used as follows.

[0029] Level A is a normal state in which no temperature abnormality has occurred in the connection 5. Level B is a state in which it is necessary to check the temperature trend of the connection 5 at the next inspection. Level C is a state in which it is necessary to reduce the frequency of inspection of the connection 5. For example, when it is judged to be at level C, the frequency of inspection of the judged connection 5 may be increased from once a year to three times a year. Level D is a state in which it is highly likely that the monitored object 3 is not connected to the electric wire 4, and it is necessary to shut down the power equipment 2 and inspect it.

[0030] Here, an example of a method for setting the level of temperature anomaly, which is a label used in the machine learning device 10, will be described.

[0031] For example, the temperature anomaly level may be set by comparing input data such as a thermal image of the connection part 5 for which the temperature anomaly level is set with input data such as a thermal image of the same connection part 5 at multiple past time points. Specifically, the relationship between the temperature data and air temperature data of the same connection part 5 at multiple past time points is graphed, and a reference line such as an approximation curve is obtained. Then, as the value of the temperature data obtained from the thermal image relative to the air temperature data deviates from the reference line, the temperature anomaly level may be increased. At this time, the temperature anomaly level may be corrected taking into consideration the current data flowing through the connection part 5, the humidity data at the time of acquiring the thermal image, and the sunshine data. For example, when the current value flowing through the connection part 5 or the amount of sunshine on the connection part 5 increases, the temperature of the connection part 5 also increases. For this reason, the temperature anomaly level may be set by adjusting the temperature data of the connection part 5 obtained from the thermal image according to the value of the current data and the sunshine data, and then comparing it with the past data. Furthermore, the temperature anomaly level may be set taking into consideration the three-phase temperature difference of the connection part 5 obtained from the thermal image. For example, as the temperature difference between the connection parts 5 of the three phases having the same connection part identification number increases, the temperature anomaly level may be increased. In addition, for example, the temperature abnormality level may be set by comparing with thermal images, current data, weather data, etc. of the connection 5 between the monitored object 3 and the electric wire 4 in other similar power equipment 2, rather than with past data.

[0032] The storage unit 13 stores the thermal image, the connection part identification number, the current data, and the weather data acquired by the input data acquisition unit 11 as input data, and the temperature anomaly level acquired by the label acquisition unit 12 as a label. Specifically, the storage unit 13 stores the thermal image, the connection part identification number of the connection part 5 that is the subject of the thermal image, the current data flowing through the connection part 5, and the weather data at the time of acquisition of the thermal image as a set of data linked to each other. The storage unit 13 then stores the input data, which is a set of data including the thermal image, the connection part identification number, the current data, and the weather data that are linked to each other, and the label, which is the temperature anomaly level for the input data, as a set of teacher data by linking them to each other. The storage unit 13 also stores the learning model constructed by the learning model construction unit 14.

[0033] The learning model construction unit 14 constructs a learning model for determining the level of temperature anomaly of the connection part 5 included as a subject in the thermal image by performing supervised learning using pairs of the input data and the labels stored in the storage unit 13 as teacher data. Then, the learning model construction unit 14 transmits the constructed learning model to the temperature anomaly determination device 20.

[0034] The learning model construction unit 14 can be realized, for example, by using a Support Vector Machine (hereinafter also referred to as SVM).

[0035] The temperature anomaly determination device 20 will be described with reference to Fig. 6. Fig. 6 is a functional block diagram of the temperature anomaly determination device 20. The temperature anomaly determination device 20 includes a control unit 21, a storage unit 22, a communication unit 23, and a display unit 24.

[0036] The control unit 21 is a part that controls the entire temperature anomaly determination device 20, and realizes various functions in this embodiment by appropriately reading and executing various programs from a storage area such as a ROM, RAM, flash memory, or hard disk (HDD). The control unit 21 may be a CPU. The control unit 21 includes a thermal image acquisition unit 210, a service number acquisition unit 211, a current data acquisition unit 212, a weather data acquisition unit 213, a temperature identification unit 214, a three-phase temperature difference identification unit 215, a temperature anomaly determination unit 216, and an output unit 217.

[0037] The thermal image acquisition unit 210 acquires a new thermal image including the connection part 5 as a subject, separately from the thermal image including the connection part 5 as a subject included in the teacher data used in machine learning by the machine learning device 10. For example, an operator may operate an input device such as a keyboard or a mouse provided in the temperature anomaly determination device 20 to cause the temperature anomaly determination device 20 to read a new photo including the connection part 5 as a subject, and then the thermal image acquisition unit 210 may acquire the new photo that has been read.

[0038] The identification number acquisition unit 211 acquires the connection part identification number of the connection part 5 that is the subject shown in the thermal image acquired by the thermal image acquisition unit 210.

[0039] The current data acquiring unit 212 acquires current data which is the value of a current flowing through the connection unit 5. The current data is acquired from, for example, a monitoring system which monitors the current flowing through the power equipment 2.

[0040] The weather data acquisition unit 213 includes a temperature data acquisition unit 213a, a humidity data acquisition unit 213b, and an insunshine data acquisition unit 213c.

[0041] The air temperature data acquiring unit 213a acquires the air temperature at the time of acquiring the thermal image acquired by the thermal image acquiring unit 210. Specifically, the air temperature data acquiring unit 213a acquires the air temperature around the connection unit 5 included in the thermal image at the time of acquiring the thermal image as air temperature data. The air temperature data may be acquired by a thermometer, a temperature sensor, or the like installed around the connection unit 5.

[0042] The humidity data acquiring unit 213b acquires the humidity at the time of acquiring the thermal image acquired by the thermal image acquiring unit 210. Specifically, the humidity data acquiring unit 213b acquires the humidity around the connection unit 5 contained in the thermal image at the time of acquiring the thermal image as humidity data. The humidity data may be acquired by a hygrometer, a humidity sensor, or the like installed around the connection unit 5.

[0043] The sunshine data acquisition unit 213c acquires sunshine data, which is the amount of sunshine irradiated on the connection unit 5. The sunshine data is acquired from a sunshine meter or the like installed around the connection unit 5.

[0044] The temperature identifying section 214 identifies the temperature of the connection section 5 reflected in the thermal image acquired by the thermal image acquiring section 210 .

[0045] The three-phase temperature difference identifying unit 215 identifies the temperature difference between the three phases of the connection portion 5 shown in the thermal image from the temperatures of the connection portion 5 for the three phases identified by the temperature identifying unit 214.

[0046] The temperature abnormality determination unit 216 determines the level of temperature abnormality of the connection part 5, which is the subject of the thermal image acquired by the thermal image acquisition unit 210, based on the learning model.

[0047] The output unit 217 outputs the level of the temperature abnormality determined by the temperature abnormality determination unit 216 to the display unit 24. In addition to the level of the temperature abnormality, the output unit 217 may also output to the display unit 24 the temperature of the connection unit 5 reflected in the thermal image identified by the temperature identification unit 214 and the temperature difference between the three phases of the connection unit 5 identified by the three-phase temperature difference identification unit 215.

[0048] The storage unit 22 stores the learning model acquired from the machine learning device 10. The storage unit 22 may store, as a set of input data, the thermal image acquired by the thermal image acquisition unit 210, meteorological data and current data related to the thermal image, a connection part identification number, and the like, and may store, as a label, the level of temperature anomaly determined for the input data.

[0049] The communication unit 23 transmits and receives data to and from the machine learning device 10. For example, the temperature anomaly determination system 1 can transmit a learning model constructed by the learning model construction unit 14 from the machine learning device 10 to the temperature anomaly determination device 20 via the communication unit 23. Conversely, the temperature anomaly determination system 1 can transmit a new thermal image acquired by the thermal image acquisition unit 210, data such as meteorological data, current data, and connection part identification number related to the thermal image, and the level of the determined temperature anomaly from the temperature anomaly determination device 20 to the machine learning device 10 via the communication unit 23.

[0050] The display unit 24 is a monitor that displays the result of the determination of the temperature abnormality level output from the output unit 217 of the control unit 21, the temperature of the connection part 5 reflected in the thermal image, the temperature difference between the three phases, and the like.

[0051] Next, a description will be given of the operation during machine learning in the temperature anomaly determination system 1 according to this embodiment. Fig. 6 is a flowchart showing the operation of the machine learning device 10 during machine learning.

[0052] In step S11, the input data acquisition unit 11 of the machine learning device 10 acquires, as input data, a thermal image, a connection part identification number linked to the thermal image, current data, and weather data.

[0053] In step S12, the label acquiring unit 12 of the machine learning device 10 acquires the level of the temperature abnormality of the connection unit 5 as a label.

[0054] In step S13, the learning model construction unit 14 of the machine learning device 10 accepts a pair of input data and a label as training data.

[0055] In step S14, the learning model construction unit 14 of the machine learning device 10 executes machine learning using this teacher data.

[0056] In step S15, the learning model construction unit 14 determines whether to end the machine learning or to repeat the machine learning. If the learning model construction unit 14 determines to repeat the machine learning (No in step S15), the process returns to step S11. Then, the machine learning device 10 repeats the same operation. On the other hand, if the learning model construction unit 14 determines to end the machine learning (Yes in step S15), the process proceeds to step S16. Note that the condition for ending the machine learning can be determined arbitrarily. For example, the machine learning may be ended when the machine learning has been repeated a predetermined number of times.

[0057] In step S16, the machine learning device 10 transmits the learning model constructed by the machine learning up to that point to the temperature anomaly determination device 20 via a network or the like.

[0058] Furthermore, the storage unit 13 of the machine learning device 10 stores this learning model. Thereby, when the learning model is requested by the temperature anomaly determination device 20, the learning model can be transmitted to the temperature anomaly determination device 20. Furthermore, when new teacher data is acquired, further machine learning can be performed on the learning model.

[0059] Next, an example of a process of temperature anomaly determination at the connection portion 5 between the monitored object 3 and the electric wire 4 by the temperature anomaly determination device 20 will be described with reference to Fig. 8. Fig. 8 is a flowchart showing the flow of the temperature anomaly determination process by the temperature anomaly determination device 20.

[0060] In step S21, the thermal image acquiring unit 210 acquires a new thermal image including, as a subject, the connection parts 5 between the monitored object 3 and the electric wires 4. The thermal image acquired by the thermal image acquiring unit 210 at this time includes, as a subject, the connection parts 5 between the monitored object 3 and all three-phase electric wires 4.

[0061] In step S22, the connection part identification number, current data, and weather data related to the thermal image acquired in step S21 are acquired. Specifically, the identification number acquisition unit 211 acquires the connection part identification number of the connection part 5 which is the subject of the thermal image acquired in step S21. In addition, the current data acquisition unit 212 acquires current data flowing through the connection part 5 which is the subject of the thermal image acquired in step S21. In addition, the weather data acquisition unit 213 acquires temperature data, humidity data, and sunlight data irradiated on the connection part 5 around the connection part 5 which is the subject of the thermal image acquired in step S21.

[0062] In step S23, the temperature identifying unit 214 identifies the temperatures of the connection parts 5 for three phases reflected in the thermal image acquired in step S21.

[0063] In step S24, the three-phase temperature difference identifying unit 215 identifies the temperature difference between the three phases of the connection portion 5 from the temperatures of the connection portion 5 for the three phases identified in step S23.

[0064] In step S25, the temperature anomaly determination unit 216 determines the level of temperature anomaly based on a set of data including the thermal image acquired in step S21, the connection part identification number acquired in step S22, the current data, and the weather data, and the learning model transmitted from the machine learning device 10.

[0065] In step S26, the output unit 217 outputs the temperatures of the three phases of the connection part 5 identified in step S23, the temperature difference between the three phases identified in step S24, and the level of temperature abnormality determined in step S25 to the display unit 24. As a result, the temperature of the connection part 5 reflected in the new thermal image acquired by the thermal image acquisition unit 210, the temperature difference between the three phases, and the level of temperature abnormality are displayed on the display unit 24.

[0066] The machine learning device 10 or the temperature anomaly determination device 20 of the temperature anomaly determination system 1 according to the present embodiment described above provides the following advantages.

[0067] The machine learning device 10 of this embodiment is a machine learning device 10 that constructs a learning model for determining a temperature abnormality at the connection 5 between the monitored object 3 and the electric wire 4, and includes an input data acquisition unit 11 that acquires a thermal image including the connection 5 as a subject as input data, a label acquisition unit 12 that acquires the level of temperature abnormality at the connection 5 as a label, and a learning model construction unit 14 that constructs a learning model for determining the level of temperature abnormality at the connection 5 included as a subject in the thermal image by performing supervised learning using a pair of the input data and the label as teacher data.

[0068] This makes it possible to construct a learning model for more efficiently determining the level of temperature abnormality at the connection 5 between the monitored object 3 and the electric wire 4. Therefore, by using the learning model, even when there are a large number of connections 5, it is possible to more efficiently determine which connections 5 may have a connection abnormality or are possibly deteriorated.

[0069] In the machine learning device 10 according to this embodiment, the input data acquisition unit 11 further acquires, as input data, the value of the current flowing through the connection unit 5 and meteorological information.

[0070] This makes it possible to construct a learning model with higher accuracy that can determine the level of temperature abnormality at connection part 5 by taking into account multiple factors such as weather information such as outside temperature and humidity that affect the temperature of connection part 5, and the current value flowing through connection part 5.

[0071] In the machine learning device 10 of this embodiment, the monitored object 3 is connected to a three-phase AC electric wire 4, and the thermal image acquired by the input data acquisition unit 11 includes the connection points 5 between the monitored object 3 and all three phases of the electric wires 4 as subjects.

[0072] As a result, since the thermal image of the three-phase connection portion 5 is used as input data, a learning model can be constructed that can determine the level of temperature abnormality with higher determination accuracy by taking into account the temperature difference of the three-phase connection portion 5.

[0073] Furthermore, the temperature anomaly determination device 20 according to this embodiment is a temperature anomaly determination device 20 that uses a learning model constructed by the machine learning device 10, and includes a thermal image acquisition unit 210 that acquires a new thermal image including the connection portion 5 of the monitored object 3 and the electric wire 4 as a subject, and a temperature anomaly determination unit 216 that determines the level of temperature anomaly at the connection portion 5 of the monitored object 3 and the electric wire 4 based on the thermal image and the learning model.

[0074] As a result, the level of temperature abnormality is determined by inputting a new thermal image into the learning model of this embodiment, so that connections 5 that may have a connection abnormality or are possibly deteriorated can be determined more efficiently.

[0075] Although the embodiment of the present invention has been described above, the present invention is not limited to the above embodiment and can be modified as appropriate.

[0076] In the above embodiment, the machine learning device 10 performed supervised learning by acquiring thermal images, current data, weather information, and a connection part recognition number as input data, but the device may be configured to perform supervised learning by using only thermal images as input data, without acquiring current data, weather information, or a connection part recognition number.

[0077] In the above embodiment, the machine learning device 10 performs supervised learning using a thermal image including, as an object of observation, the connection parts 5 between the monitored object 3 and all three-phase electric wires 4 as input data, but the connection parts 5 between all three-phase electric wires 4 may not be included as objects of observation. For example, the machine learning device 10 may perform supervised learning using, as input data, a thermal image including, as an object of observation, the connection parts 5 between the monitored object 3 and all three-phase electric wires 4. [Explanation of symbols]

[0078] 3. Objects of surveillance 4 electric wire 5,5a,5b,5c,5d,5e,5f connection part 10 Machine Learning Device 11 Input data acquisition section 12 Label Acquisition Section 14 Learning Model Construction Department 20 Temperature abnormality determination device

Claims

1. A machine learning device that constructs a learning model for determining a temperature anomaly at a connection between a monitoring object and an electric wire, an input data acquisition unit that acquires a thermal image including the connection portion as an object as input data; a label acquiring unit that acquires a level of the temperature abnormality of the connection portion as a label; a learning model construction unit that constructs a learning model for determining a level of temperature anomaly at the connection portion included as a subject in the thermal image by performing supervised learning using a pair of the input data and the label as teacher data, The input data acquisition unit is a machine learning device that further acquires, as the input data, a current value flowing through the connection portion and weather information.

2. The object to be monitored is connected to a three-phase AC electric line, The machine learning device according to claim 1 , wherein the thermal image acquired by the input data acquisition unit includes, as a subject, a connection portion between the monitored object and all three-phase electric wires.

3. A temperature anomaly determination device using the learning model constructed by the machine learning device according to claim 1 or 2, a thermal image acquisition unit for acquiring a new thermal image including a connection portion between the monitored object and the electric wire as an object to be monitored; A temperature anomaly determination device comprising: a temperature anomaly determination unit that determines the level of temperature anomaly at the connection between the monitored object and the electric wire based on the thermal image and the learning model.

Images