Electrical equipment abnormality detection system

Abstract

To provide a failure detection system for electrical equipment capable of appropriately detecting the occurrence of an abnormality.SOLUTION: A failure detection system (1) for electrical equipment comprises: a notification unit that indicates an abnormality on the basis of a distance or similarity between detection signal data which is based on an output signal of a detection unit (10) that monitors an atmosphere gas inside a housing of the electrical equipment and teacher data. The teacher data and the detection signal data are data corresponding the respective outputs of a plurality of odor sensors (12) of the detection unit.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to an abnormality detection system for an electrical device. [Background technology]

[0002] Anomaly detection systems have been proposed to detect abnormalities such as abnormal heat generation and burnout in electrical equipment, such as power facility equipment (power generation / transformation / receiving facility equipment) for power generation, transformation, receiving, etc. One such anomaly detection system has been devised in which an odor sensor is installed inside the housing of the power generation / transformation / receiving facility equipment, and the odor generated by abnormal heat generation, burnout, abnormal discharge, etc. is detected, thereby enabling the occurrence of these abnormalities to be detected. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Patent No. 2755802 [Patent Document 2] Patent No. 3729526 [Patent Document 3] Patent No. 3838461 [Patent Document 4] Japanese Patent Application Laid-Open No. 2008-76209 Summary of the Invention [Problem to be solved by the invention]

[0004] The odor sensor outputs a sensor output corresponding to the amount of the chemical substance that produces the odor. In these conventional technologies, the presence or absence of an abnormality is determined based on whether the magnitude of the sensor output exceeds a predetermined threshold.

[0005] However, the amount of chemical substances (smell-causing substances) that reach the odor sensor when an abnormality occurs varies greatly depending on the relative position of the abnormality location inside the housing of the electrical device and the installation position of the detection unit equipped with the odor sensor, the flow of atmospheric gas in the space inside the housing, the state of ventilation, etc. Therefore, it is not easy to determine a threshold value for the sensor output that can accurately determine whether or not an abnormality exists.

[0006] A technology has also been devised that determines the presence or absence of an abnormality not simply by the absolute value of the sensor output, but by installing an odor sensor in an external space where the above abnormalities cannot occur, and determining the relative value using the sensor output as a reference. However, even in this case, for the same reasons as above, it is not easy to determine a threshold value that can determine the presence or absence of an abnormality.

[0007] Furthermore, it is not easy to set up a space that can serve as an appropriate reference. One aspect of the present invention aims to realize an abnormality detection system for electrical equipment that can appropriately detect the occurrence of an abnormality in an electrical equipment by mitigating the influence of the relative relationship between the installation location of a detection unit and the location where an abnormality occurs. [Means for solving the problem]

[0008] In order to solve the above problems, one embodiment of the present invention provides an abnormality detection system for electrical equipment, comprising: a detection unit that monitors the atmospheric gas inside the housing of the electrical equipment; a recording unit that stores teacher data; a signal processing unit that outputs detection signal data based on the output signal of the detection unit; a comparison unit that calculates the distance or similarity between the detection signal data and the teacher data; and an alarm unit that alerts to an abnormality based on the distance or similarity, wherein the detection unit is provided with two or more different odor sensors, and the teacher data and the detection signal data are data corresponding to the outputs of the multiple odor sensors. [Effects of the Invention]

[0009] According to an electrical device abnormality detection system according to one aspect of the present invention, the influence of the relative relationship between the installation location of the detection unit and the location where the abnormality occurs is mitigated, making it possible to appropriately detect the occurrence of an abnormality in the electrical device. [Brief explanation of the drawings]

[0010] [Figure 1] 1 is a diagram illustrating a configuration of an anomaly detection system according to a first embodiment of the present invention. [Figure 2] 1 is a schematic diagram showing the outer shape of a detection unit in an abnormality detection system according to a first embodiment of the present invention. [Figure 3] 1 is a diagram showing an installation state of an anomaly detection system in an electrical device according to a first embodiment of the present invention. [Figure 4] FIG. 3 is a diagram showing an example of detection signal data in the form of a radar chart in the first embodiment of the present invention. [Figure 5] FIG. 2 is a diagram showing an example of training data in the form of a radar chart in the first embodiment of the present invention. [Figure 6] FIG. 4 is a schematic diagram showing the outer shape of a detection unit in a modified example of the first embodiment of the present invention. [Figure 7] FIG. 10 is a diagram showing an installation state of an anomaly detection system in an electrical device according to a fourth embodiment of the present invention. [Figure 8] 1 is a time chart showing changes in the sensor output of each odor sensor of the detection unit, along with the humidity of the ambient gas to be monitored. [Figure 9] 10 is a time chart showing the change when the sensor output of each odor sensor of the detection unit is corrected, along with the humidity of the ambient gas to be monitored. [Figure 10] 4 is a time chart showing the sensor output of each odor sensor of the detection unit. [Figure 11] 10 is a time chart when correction is made to the sensor output of each odor sensor of the detection unit. [Figure 12] FIG. 16 is a diagram illustrating the configuration of an anomaly detection system according to a tenth embodiment of the present invention. [Figure 13]FIG. 22 is a diagram showing an example of detection signal data in the form of a radar chart in the eleventh embodiment of the present invention. [Figure 14] FIG. 22 is a diagram showing an example of detection signal data in the form of a radar chart in the eleventh embodiment of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0011] [Embodiment 1] <Configuration of anomaly detection system> An embodiment of the present invention will be described in detail below with reference to Figures 1 to 6. Figure 1 is a block diagram showing an anomaly detection system 1 according to embodiment 1. The anomaly detection system 1 includes a detection unit 10 and an information processing unit 20.

[0012] The detection unit 10 is a device for monitoring the atmospheric gas inside the housing of an electrical device. FIG. 2 is a schematic diagram showing the external shape of the detection unit 10. The detection unit 10 is provided with a plurality of odor sensors 12 of at least two different types. In addition to the plurality of odor sensors 12, the detection unit 10 is also provided with a printed circuit board 11 and a detection circuit 13. As shown in FIG. 2, the main body of the detection unit 10 is configured such that the plurality of odor sensors 12 and the detection circuit 13 are mounted on the printed circuit board 11.

[0013] The detection unit 10 may further include other odor sensors that are not included in the plurality of odor sensors 12. The odor sensors 12 may be mounted on the printed circuit board 11 by soldering. Alternatively, the odor sensors 12 may be removably inserted into a sensor connector provided on the printed circuit board for each odor sensor 12, allowing for easy replacement.

[0014] The multiple odor sensors 12 of the detection unit 10 are each a different type of sensor, each sensitive to a different odor (odor-causing chemical substance) and generating a sensor output according to the amount of sensitivity. The odor sensor 12 is composed of an oscillator and an adsorption film provided on the oscillator that adsorbs a specific odor-causing chemical substance. In such an odor sensor 12, the amount of the odor-causing chemical substance can be detected by detecting in the detection circuit 13 a change in the resonant frequency of the oscillator that occurs when a specific odor-causing chemical substance (e.g., a gas) adheres to the adsorption film.

[0015] The arrangement of the odor sensors shown in Figure 2 is a schematic example, and is not limited to the arrangement or number of odor sensors shown in the figure. In the present invention, the detection unit 10 may include two or more types of odor sensors 12. However, when displaying data using a radar chart, it is preferable that the detection unit 10 include three or more types of odor sensors 12. In the following description of Embodiment 1, a case where there are six types of odor sensors 12 is exemplified.

[0016] The detection circuit 13 controls and monitors the odor sensors 12, and transmits information including sensor output corresponding to the amount of odor-causing substances sensed by each odor sensor 12 to the information processing unit 20 as an output signal Sd from the detection unit 10. The information processing unit 20 has the following functional blocks: a recording unit 21, a signal processing unit 22, a comparison unit 23, and an alarm unit 24. The recording unit 21 is a functional block that stores training data Pr. The signal processing unit 22 is a functional block that receives the output signal Sd from the detection unit 10 and outputs detection signal data Pd based on it.

[0017] The comparison unit 23 is a functional block that calculates the distance or similarity Dc between the detection signal data Pd output by the signal processing unit 22 and the teacher data Pr held by the recording unit 21. The notification unit 24 is a functional block that determines whether or not an abnormality exists in the electrical equipment monitored by the abnormality detection system 1, based on the distance or similarity Dc calculated by the comparison unit 23, and notifies the user of the presence or absence of an abnormality.

[0018] <Example of anomaly detection system installation> 3 is a schematic diagram showing an example of installation of the anomaly detection system 1 in an electrical device. In the first embodiment, the detection unit 10 of the anomaly detection system 1 is installed in the internal space of the electrical device housing 9. Preferably, the detection unit 10 is installed in the upper part of the internal space of the electrical device housing 9. The information processing unit 20 of the anomaly detection system 1 is installed in the electrical device housing 9 as a panel, part of which can be seen from the outside. The detection unit 10 and the information processing unit 20 are connected as appropriate by signal lines for transmitting the output signal Sd and the like.

[0019] Specific examples of electrical equipment include high-voltage distribution panels such as cubicles used in power stations and substations, factories, etc., gas insulated switches (GIS), transformers, etc., but the equipment may also be other power generation and transformer power receiving equipment, especially extra-high voltage equipment, or other electrical equipment.

[0020] The installation method of the anomaly detection system 1 shown in Fig. 3 is an example of a specific configuration in embodiment 1. As another example, the information processing unit 20 may be installed inside the electrical equipment housing 9 in a state where it is not visible from the outside. Alternatively, the information processing unit 20 may be installed outside the electrical equipment housing 9. When the electrical equipment is a gas-insulated switchgear, the detection unit 10 may be installed inside its operating mechanism or a housing in which gas is sealed, or in gas piping routed around an inlet / outlet valve, etc.

[0021] <Operation of the detection unit and signal processing unit> Fig. 4 is a diagram showing an example of a radar chart display of detection signal data Pd output by signal processing unit 22 based on output signal Sd from detector 10 installed in electrical device housing 9 in various states of the electrical device. Fig. 4 shows detection signal data 4a, detection signal data 4b, detection signal data 4c, and detection signal data 4d, which are detection signal data Pd, for the cases where the state of the electrical device is abnormal 1, abnormal 2, abnormal 3, and normal, respectively.

[0022] Here, an abnormal state refers to a state in which an abnormal odor is generated due to, for example, carbonization of the surface of molded parts or carbonization of wire coatings due to partial discharge or abnormal overheating, abnormal overheating of transformers, switches, molded equipment, etc., or abnormal overheating of the distribution lines or other specific parts of such electrical equipment.

[0023] Furthermore, a normal state refers to a state in which the electrical device is operating normally and no abnormalities have occurred. When an abnormality accompanied by an odd odor occurs inside the electrical device housing 9, the intensity of the sensor output of each of the different types of odor sensors 12 that responds to the odor-causing substance generated at the time of the abnormality increases.

[0024] 4, the signal processing unit 22 outputs data including the magnitude of each sensor output as detection signal data Pd based on the sensor outputs of the six odor sensors 12 provided in the detection unit 10. In this case, the signal processing unit 22 normalizes the data as a set including the magnitude of each sensor output for each measurement to generate the detection signal data Pd.

[0025] When such normalized detection signal data Pd is displayed as, for example, a radar chart, the distance from the center of the data plotted on each axis (channels CH1 to CH6) that is farthest from the center of the radar chart is always constant. In other words, the radar chart is drawn focusing only on the shape. In other words, the signal processing unit 22 generates detection signal data Pd focusing on the relative relationship between the magnitudes of the sensor outputs, rather than the absolute values ​​of the sensor outputs.

[0026] When some kind of abnormality occurs in an electrical device, the odor-causing chemical substance that is generated varies depending on the type of abnormality. Therefore, as shown in Figure 4, the sensitivity of each odor sensor 12 differs depending on the type of abnormality (or normal state), and the signal processing unit 22 outputs detection signal data Pd that have different shapes when represented on a radar chart.

[0027] <Generating training data> Next, there will be described a method for generating the training data Pr stored in the recording unit 21 of the anomaly detection system 1. The manufacturer of the anomaly detection system 1 experimentally generates a specific anomaly 1 that is expected to occur during operation of the electrical device.

[0028] Then, the signal processing unit 22 generates detection signal data Pd based on the output signal Sd of the detection unit 10 at that time. At this time, sensing is performed for about one minute at a cycle of about 5 to 10 minutes, for example, to generate a large number of detection signal data Pd. Here, the number of detection signal data Pd to be generated is preferably about 10 to 100, for example, but may be a larger number.

[0029] Similarly, sets of detection signal data Pd are generated for each of the other possible abnormalities (abnormality 2, abnormality 3, ...). Furthermore, the signal processing unit 22 also generates a large number of detection signal data Pd based on the output signal Sd of the detection unit 10 during normal operation when no abnormality has occurred in the electrical equipment.

[0030] Each set of the thus acquired detection signal data Pd for each specific abnormal and normal state is subjected to machine learning to generate patterned detection signal data that serve as training data Pr for each specific abnormal and normal state. Here, the patterned detection signal data refers to typical detection signal data for a specific case.

[0031] A specific method that can generate such training data Pr is, for example, the well-known support vector machine method. Note that it is also preferable to perform a process to remove noise from each of the acquired detection signal data Pd before machine learning.

[0032] A specific example of the teacher data Pr thus obtained is shown in the radar chart of Fig. 5. Here, machine learning was performed for four cases: abnormality 1, abnormality 2, abnormality 3, and normality. The teacher data Pr thus generated, which is patterned detection signal data for each abnormality and normality, is stored in the recording unit 21 of the anomaly detection system 1.

[0033] <Monitoring of electrical equipment by anomaly detection system> The following describes the operation of the anomaly detection system 1 to monitor the occurrence of an anomaly in an electrical device. A plurality of different types of odor sensors 12 provided in the detection unit 10 respond to odors (odor-causing chemical substances) in the atmospheric gas inside the electrical device housing 9 according to their respective characteristics. The detection circuit 13 of the detection unit 10 detects the sensor output of each odor sensor 12 and transmits an output signal Sd containing this information to the signal processing unit 22 of the information processing unit 20.

[0034] The signal processing unit 22 generates detection signal data Pd based on the sensor output information of the six types of odor sensors 12 in the output signal Sd. A specific example of the detection signal data Pd is shown in the six-axis radar chart in Fig. 4. When abnormality 1, abnormality 2, or abnormality 3 occurs within the electrical device casing 9, the signal processing unit 22 generates detection signal data Pd such as detection signal data 4a, detection signal data 4b, and detection signal data 4c, respectively.

[0035] When the electrical equipment is operating normally, detection signal data Pd such as detection signal data 4d is generated by the signal processing unit 22. The comparison unit 23 compares the detection signal data Pd generated by the signal processing unit 22 with multiple pieces of teacher data Pr recorded in the recording unit 21, and calculates the distance or similarity Dc for each piece.

[0036] For example, if the detection signal data Pd is the detection signal data 4a in Fig. 4, the comparison unit 23 calculates the distance or similarity Dc between each piece of teacher data Pr shown in Fig. 5 and the detection signal data 4a, which are held by the recording unit 21. The distance or similarity Dc with the teacher data 5a was calculated to be 82.6%, and the distances or similarities Dc with the other teacher data were significantly lower than that.

[0037] Next, the notification unit 24 determines whether or not an abnormality has occurred based on each distance or similarity Dc calculated by the comparison unit 23. In this case, the teacher data 5a showing the largest distance or similarity Dc is patterned detection signal data indicating abnormality 1, so the notification unit 24 determines that an abnormality has occurred in the electrical equipment. Furthermore, if abnormality 1 has occurred in the electrical equipment, the notification unit 24 determines by identifying the type of abnormality.

[0038] The notification unit 24 notifies the manager of the electrical equipment of the occurrence of an abnormality and / or the type of the abnormality that has occurred (abnormality 1). The notification may be made by a display, a warning light, a buzzer, a voice output, or the like that is appropriately provided in the notification unit 24. Alternatively, the manager may be notified via an appropriate communication means.

[0039] When the detection signal data Pd generated by the signal processing unit 22 was the detection signal data 4b of Figure 4, the distance or similarity Dc calculated by the comparison unit 23 was 89.1% with the training data 5b indicating abnormality 2 in Figure 5, and the distance or similarity Dc with other training data was significantly lower.

[0040] When the detection signal data Pd generated by the signal processing unit 22 was the detection signal data 4c in Fig. 4, the distance or similarity Dc calculated by the comparison unit 23 was 52.0% with the teacher data 5c indicating abnormality 3 in Fig. 5, and the distances or similarities Dc with the other teacher data were significantly lower. In these cases, the notification unit 24 was able to determine that abnormality 2 or abnormality 3 had occurred in the electrical equipment, respectively, and operated in the same manner as above in accordance with the determination results.

[0041] When the detection signal data Pd generated by the signal processing unit 22 was the detection signal data 4d of FIG. 4, the distance or similarity Dc calculated by the comparison unit 23 was 83.6% with the teacher data 5d indicating normality in FIG. 5, and the distances or similarities Dc with the other teacher data were significantly lower. In this case, the teacher data 5a showing the largest distance or similarity Dc is patterned detection signal data indicating normality, so the notification unit 24 determined that the electrical equipment is in a normal state. As a result, the notification unit 24 does not notify the administrator of the abnormality.

[0042] The calculation of the distance or similarity Dc between the detection signal data Pd and the training data Pr in the comparison unit 23 is performed by, for example, a known pattern recognition method.

[0043] <Actions and Effects> The anomaly detection system 1 is capable of detecting various anomalies in electrical equipment because it is equipped with at least two or more types of odor sensors 12. Furthermore, the anomaly detection system 1 detects anomalies based on detection signal data Pd (corresponding to the shape of the radar chart) that focuses on the relative relationship between the magnitudes of the outputs of at least two or more sensors, and does not depend on the absolute values ​​of the sensor outputs of the odor sensors 12.

[0044] Therefore, the possibility of detecting an abnormality is prevented from being affected by the relative relationship between the location of the abnormality inside the housing of the electrical device and the installation position of the detection unit equipped with the odor sensor, the flow of atmospheric gas in the internal space of the electrical device housing 9, the state of ventilation, etc. Furthermore, this eliminates the difficulty of determining a threshold value for the magnitude of the sensor output for determining an abnormality, as was the case with conventional technology.

[0045] According to the anomaly detection system 1, anomalies are detected based on detection signal data Pd (corresponding to the shape of the radar chart) that focuses on the relative relationship between the magnitudes of the outputs of at least two or more sensors, and therefore, compared to technologies that rely on information from a specific odor sensor 12 for detection, the occurrence of false detections is reduced.

[0046] According to the anomaly detection system 1, anomalies are detected based on the distance or similarity of the detection signal data Pd to the training data Pr, so by preparing training data Pr for various anomalies, it becomes possible to easily detect and distinguish between various types of anomalies.

[0047] <Modification> This modification is an example in which the configuration of the detection unit 10 shown in Fig. 2 is modified. For ease of explanation, members having the same functions as those described above are denoted by the same reference numerals. Fig. 6 is a schematic diagram showing the outer shape of the detection unit 10 in this modification.

[0048] The detection unit 10 of this modified example is provided with two separate printed circuit boards, printed circuit board 11a and printed circuit board 11b. A large number of odor sensors 12 are arranged in a row on the upper surface of printed circuit board 11a. A connector 14a is arranged in one or more locations on the lower surface of printed circuit board 11a. A connector 14b is arranged in one or more locations on the upper surface of printed circuit board 11b, and a detection circuit 13 is also arranged thereon.

[0049] Printed circuit boards 11a and 11b are fixed together by connecting connectors 14a and 14b. In this manner, in this modification, multiple printed circuit boards are stacked by the connectors. By configuring the detection unit 10 with multiple stacked printed circuit boards, it becomes easy to mount a large number of odor sensors 12.

[0050] <Additional Notes> In embodiment 1, the odor sensor 12 is exemplified as a vibrator-type sensor, but the odor sensor used in the present invention may also be an odor sensor that senses odors (odor-causing chemical substances) using other principles, such as a semiconductor-type sensor.

[0051] The odor sensor 12 used in the present invention may be of a type that is primarily sensitive to only a specific odor-causing chemical substance (for example, a specific gas), but may also be of a type that is sensitive to multiple odor-causing chemical substances (for example, multiple gases).

[0052] To deal with the estimated 40,000 types of odors using only the former type of odor sensors would require a huge number of odor sensors 12. Therefore, providing the detection unit 10 with the latter type of odor sensors 12 is effective in detecting a variety of chemical substances while limiting the number of types of odor sensors 12.

[0053] For the latter type of odor sensor, it is preferable to select one that has an appropriate detection range and sensitivity, such as high sensitivity to certain odor-causing chemicals and low sensitivity to some other odor-causing chemicals.

[0054] Alternatively, the detection unit 10 may be configured to include an odor sensor that is sensitive only to gases such as ozone (O3) or hydrogen sulfide (HS), respectively, as the former type of odor sensor. By including such a specific odor sensor 12, the accuracy of detecting abnormalities in specific electrical devices can be improved.

[0055] For example, ozone is generated by poor contact (partial discharge or spark discharge) in the high-voltage or low-voltage sections of electrical equipment (hydrogen sulfide is generated in sulfur hexafluoride (SF6) gas-insulated equipment), and hydrogen sulfide, a type of corrosive gas generated in petrochemical plants, sewage treatment plants, and various types of wastewater treatment plants, enters the interior of electrical equipment, causing corrosion of the contact parts and increasing contact resistance.

[0056] In the first embodiment, six types of odor sensors 12 are exemplified. However, the number of types of odor sensors 12 is not limited to six, and may be at least two or more. However, when a radar chart is used as the data display format, it is desirable for the radar chart to have three axes, or even four or more axes. Note that the more types of odor sensors 12 there are, the more diverse anomalies can be handled and the more detailed anomaly discrimination becomes possible, which is desirable.

[0057] In the present invention, the teacher data Pr stored in the recording unit 21 does not need to be data for all of the odor sensors included in the detection unit 10, but may be data for two or more types of odor sensors appropriately selected from among them. In this case, the detection signal data Pd compared with the teacher data Pr will naturally be composed of data for the same odor sensor as the teacher data Pr. In this way, different types of odor sensors 12 may be selected depending on the type of abnormality, thereby increasing the accuracy of abnormality detection.

[0058] If the electrical equipment is a gas insulated switchgear (GIS), the detection unit 10 may be installed in each gas compartment separated by an insulating spacer. In this case, the installation location of the detection unit 10 is not limited to inside a housing in which the gas is sealed, but may be in a gas piping routed around a gas charging / discharging valve or the like.

[0059] In this case, the odor sensor 12 applied in the present invention may be an odor sensor that is sensitive to the decomposition gas of an insulating fluid such as sulfur hexafluoride (SF6) sealed in a gas insulated switch (GIS), etc., and makes it possible to identify the gas compartment in which an abnormality occurs during a ground fault or short circuit accident.

[0060] Furthermore, the detection unit 10 may be provided outside the electrical device. In this case, the detection unit 10 detects air pollutants, such as nitrogen oxides (NO ), that may be present in the installation environment of the electrical device. X ) and sulfur oxides (SO X It is preferable that the electrical equipment further includes a gas sensor capable of detecting gases such as chlorine, nitrous oxide ...

[0061] In the anomaly detection system 1, the detection signal data Pd generated by the signal processing unit 22 is a set of numerical values ​​(scalars) based on the output of each output of the odor sensor 12, and the distance or similarity Dc is calculated and compared using such teacher data Pr and the detection signal data Pd of the signal processing unit 22.

[0062] However, the detection signal data Pd generated by the signal processing unit 22 may be a radar chart, or may be a data display format such as a bar graph or a histogram that allows a relative comparison of the magnitude of each sensor output from the detection unit 10. Furthermore, this data display format may be an image.

[0063] [Embodiment 2] Other embodiments of the present invention will be described below. For ease of explanation, the same reference numerals will be used to designate components having the same functions as those described in the above embodiment, and the description thereof will not be repeated.

[0064] The abnormality detection system 1 according to the second embodiment is the same as the abnormality detection system according to the first embodiment, except for the teacher data Pr used by the comparison unit 23. In the second embodiment, the comparison unit 23 uses only teacher data Pr indicating an abnormality as the teacher data Pr.

[0065] The method of generating the training data Pr indicating anomalies for the anomaly detection system 1 according to embodiment 2 is the same as the method described in embodiment 1. Specific examples of such training data Pr are training data 5a to 5c showing anomalies 1 to 3, respectively, shown in FIG.

[0066] In the second embodiment, during the monitoring operation of the electric device, the comparison unit 23 calculates the distance or similarity between the detection signal data Pd and each of the plurality of teacher data Pr. When the notification unit 24 finds that the distance or similarity relating to a specific abnormality is equal to or greater than a predetermined value (for example, equal to or greater than 50%), it determines that the abnormality has occurred in the electric device.

[0067] In this embodiment, the same effects as in the first embodiment can be obtained.

[0068] [Embodiment 3] The abnormality detection system according to the third embodiment is the same as the abnormality detection system according to the first embodiment, except for the teacher data Pr used by the comparison unit 23. In the third embodiment, the comparison unit 23 uses only teacher data Pr that indicates normality as the teacher data Pr.

[0069] The method for generating the training data Pr indicating normality for the anomaly detection system according to the third embodiment is as follows.

[0070] The manufacturer of the anomaly detection system 1 causes the signal processing unit 22 to generate detection signal data Pd based on the output signal Sd of the detection unit 10 during normal operation when no anomaly has occurred in the electrical equipment. Here, the number of detection signal data Pd is preferably, for example, about 10 to 100, but may be a larger number. From the large number of normal detection signal data Pd thus acquired, patterned detection signal data that will become training data Pr is generated.

[0071] More specifically, a large amount of detection signal data may be acquired from the electrical device for which abnormality detection is to be performed in a manufacturing plant during normal operation. Alternatively, a large amount of detection signal data may be acquired as normal data over a learning period (e.g., one week, or any other period that is deemed appropriate) immediately after the electrical device for which abnormality detection is to be performed begins operating on-site.

[0072] In the third embodiment, during monitoring of an electrical device, the comparison unit 23 calculates the distance or similarity between the detection signal data Pd and the teacher data Pr indicating normality. The notification unit 24 checks whether the distance or similarity is equal to or less than a predetermined value. If it is equal to or less than the predetermined value, the notification unit 24 determines that an abnormality has occurred in the electrical device. Otherwise, the notification unit 24 determines that the electrical device is operating normally.

[0073] In this way, in the abnormality detection system 1 according to the third embodiment, whether or not an electrical device is abnormal is determined based on whether or not the detection signal data Pd is similar to the teacher data Pr indicating normality.

[0074] The anomaly detection system 1 according to the third embodiment only needs to store in the recording unit 21 the teacher data Pr indicating normality generated by the above-described method as the teacher data Pr for detecting an anomaly in the electrical device. Therefore, in order to generate the teacher data Pr, it is not necessary to intentionally cause an abnormal state in the electrical device and acquire the detection signal data Pd. Therefore, the anomaly detection system 1 is easier to build than the above-described embodiments.

[0075] However, a group of patterned detection signal data serving as training data Pr may be generated in the same manner as in the first embodiment, and only the training data Pr showing normality may be used.

[0076] [Embodiment 4] The anomaly detection system of the fourth embodiment is almost the same as the anomaly detection system of the above-described embodiment, except that the arrangement of the detection unit 10 and the information processing unit 20 relative to the electrical device housing 9 and the configuration of the detection unit 10 differ from the example shown in FIG.

[0077] 7 is a schematic diagram showing the configuration of an anomaly detection system according to embodiment 4. In embodiment 4, the detection unit 10 and the information processing unit 20 are installed outside the electrical device housing 9. The detection unit 10 is provided with a case 15, and a printed circuit board 11 on which the odor sensor 12 shown in FIG. 2 and the like are mounted is hermetically housed in the internal space of the case 15.

[0078] A suction tube 16 is connected to the detection unit 10, and its end reaches the inside of the electrical device housing 9. A funnel-shaped suction port 161 that opens to the internal atmosphere of the electrical device housing 9 is provided at this end of the suction tube 16. A suction tube 17 is also connected to the detection unit 10, and its end is in contact with the outside air. A funnel-shaped suction port 171 that opens to the outside air is provided at this end of the suction tube 17. A purification filter 172 is connected midway through the suction tube 17.

[0079] In the abnormality detection system 1 according to the fourth embodiment, during monitoring of an electrical device, atmospheric gas inside the electrical device housing 9 is sucked into the case 15 by a pump mechanism (not shown) through the suction tube 16. In this way, the detection unit 10 monitors the atmospheric gas inside the electrical device housing 9. As in the above embodiments, the abnormality detection system 1 detects abnormalities in the electrical device.

[0080] Furthermore, in the abnormality detection system 1 according to the fourth embodiment, the following cleaning process is intermittently performed during the monitoring operation of the electrical device. In the cleaning process, a pump mechanism (not shown) draws outside air that has been purified by passing through the purification filter 172 into the case 15.

[0081] In this way, the gas inside the case 15 of the detection unit 10 is switched from the atmospheric gas inside the electrical device housing 9 to purified outside air. When the purified outside air is supplied to the odor sensor 12, the odor-causing chemical substances adhering to the adsorption film of the odor sensor 12 are gradually peeled off by the vibration of the vibrator, and cleaning is performed.

[0082] According to the fourth embodiment, cleaning of the odor sensor 12 is performed intermittently, which improves the life of the odor sensor 12 and makes it possible to extend the maintenance intervals of the anomaly detection system.

[0083] [Embodiment 5] Embodiment 5 is a modification of the anomaly detection systems according to Embodiments 1 to 3, and adds the mechanism for sucking in outside air in Embodiment 4. That is, the anomaly detection system according to Embodiment 5 is provided with a purification filter 172, and a suction tube 17 that sucks in outside air from a funnel-shaped suction port 171 is connected to the detection unit 10 shown in Fig. 3. As with Embodiment 4, the anomaly detection system according to Embodiment 5 allows cleaning of the detection unit 10.

[0084] [Embodiment 6] The anomaly detection system according to embodiment 6, which is a modification of the anomaly detection system according to embodiments 1 to 3 or embodiment 5, includes a blower fan that blows ambient atmospheric gas toward the detection unit 10. The anomaly detection system according to embodiment 6 is configured to forcibly circulate odor-causing chemical substances through the detection unit 10, thereby improving the detection sensitivity or accuracy of odor-causing chemical substances.

[0085] [Embodiment 7] The anomaly detection system according to embodiment 7 is a modified example of the anomaly detection system according to embodiments 1 to 3, 5 or 6, and has the detection unit 10 shown in FIG. 3 divided and provided in the upper and lower parts of the internal space of the electrical device housing 9.

[0086] In the anomaly detection system of embodiment 7, the accuracy of anomaly detection is improved by being configured to be able to efficiently detect both heavy odor-causing chemicals (odor-causing chemicals that rise and accumulate in the upper part of the internal space of the electrical equipment housing 9) and light odor-causing chemicals (odor-causing chemicals that fall and accumulate in the lower part of the internal space of the electrical equipment housing 9).

[0087] [Embodiment 8] In the anomaly detection system according to the eighth embodiment, which is a modification of each of the above-described embodiments, a predetermined threshold is set for each odor sensor 12. In the eighth embodiment, the notification unit 24 does not notify of an anomaly unless the predetermined threshold for each odor sensor is exceeded. The anomaly detection system according to the eighth embodiment improves the accuracy of anomaly determination, such as by preventing erroneous detection.

[0088] [Embodiment 9] A modification of each of the above-described embodiments, the anomaly detection system according to the ninth embodiment is configured to include a device for capturing images of the inside of an electrical device casing 9 (a camera for monitoring the inside, such as an infrared camera or a CCD camera), and to detect temperature rises due to overheating of the device, oil leaks, etc., by combining this with anomaly detection based on image recognition. The anomaly detection system according to the ninth embodiment improves the accuracy of anomaly detection.

[0089] [Embodiment 10] A tenth embodiment is a modification of the above-described embodiments, and in the anomaly detection system 1 according to the tenth embodiment, the information processing unit 20 performs an operation of performing preprocessing on the output signal Sd of the detection unit 10 in order to improve the accuracy of anomaly detection.

[0090] First, the principle of accuracy improvement will be explained. Figures 8 and 10 are time charts showing the output signal Sd of the detection unit 10 when a detection unit 10 equipped with 16 types of odor sensors 12 is used and the ambient gas monitored by the detection unit 10 is pure air (blank air). In Figures 8 and 10, the sensor output of each odor sensor 12 is displayed as a graph for each channel.

[0091] In addition, the humidity of the ambient gas is also shown by a dotted line in Fig. 8. Fig. 8 is a graph showing the changes in the sensor output (channel) and humidity at a specific time in arbitrary units. Fig. 10 is a graph showing the sensor output (channel) value in arbitrary units.

[0092] As shown in Figures 8 and 10, the output of each sensor fluctuates over time, even though the monitored object is pure air that does not contain odor-causing chemicals. As is clear from Figure 8, the fluctuations in each sensor output correspond to fluctuations in humidity. As shown in Figure 8, although the magnitude of the change in each sensor output when humidity changes differs depending on the adsorption film of each sensor, the tendency of the change is found to be similar.

[0093] Therefore, we attempted to cancel the influence of humidity by performing a correction according to the humidity for each sensor. For each sensor, the amount of correction according to the humidity H was expressed by the linear equation a × H + b (a and b are constants). The sensor output (channel) values ​​after this correction are shown in Figures 9 and 11. Figures 9 and 11 are graphs corresponding to Figures 8 and 10, respectively. As shown in the figures, when correction was performed, the fluctuations due to humidity were successfully canceled out. As a result, the fluctuations in sensor output over time were suppressed.

[0094] Fig. 12 is a block diagram showing the configuration of an anomaly detection system 1 according to embodiment 10. In the anomaly detection system 1 according to embodiment 10, a humidity detection unit 30 that detects the humidity of the atmosphere to be monitored is added to the configuration of the anomaly detection system shown in Fig. 1. Note that the humidity detection unit 30 may be integrated with the detection unit 10, for example, by mounting a humidity sensor on the printed circuit board 11 of the detection unit 10.

[0095] Furthermore, in the anomaly detection system 1 according to the tenth embodiment, the signal processing unit 22 performs preprocessing for each odor sensor 12 to cancel the influence of humidity as described above, based on the humidity of the monitored ambient gas detected by the humidity detection unit 30, and outputs the result as detection signal data Pd. Note that the correction based on humidity is not limited to the linear equation described above, and may be performed using other appropriate correction equations according to the characteristics of each odor sensor 12, such as a polynomial function, an exponential function, or a logarithmic function.

[0096] According to the anomaly detection system 1 of the tenth embodiment, the fluctuation in the sensor output of each odor sensor 12 due to the influence of humidity is canceled out, thereby further improving the accuracy of anomaly detection.

[0097] [Embodiment 11] An anomaly detection system 1 according to an eleventh embodiment, which is a modification of each of the above-described embodiments, is configured to be able to detect odors emanating from decaying small animals or their excrement, assuming that small animals have entered the interior of the electrical device casing 9. In the anomaly detection system 1 according to the eleventh embodiment, the number of types of odors that can be identified by the detection unit 10 is increased.

[0098] The abnormality detection system of embodiment 11 is configured to detect odor-causing chemicals such as skatole, indole, and ammonia, which are considered to be the cause of foul odors among the odors emitted by small animal feces and urine or decay, making it possible to detect the intrusion of small animals.

[0099] 13 and 14 are diagrams showing an example of a radar chart display of the detection signal data Pd output by the signal processing unit 22 of the anomaly detection system 1 according to embodiment 11 based on the output signal Sd of the detection unit 10 installed in the electrical device casing 9. As shown in Fig. 13 and 14, the detection unit 10 of the anomaly detection system 1 according to embodiment 11 includes 16 types of odor sensors 12.

[0100] Graphs 13e, 13f, and 13g show examples of detection signal data Pd output by the signal processing unit 22 based on the output signal Sd of the detection unit 10 when the detection unit 10 detects atmospheric gases containing skatole, indole, and ammonia, respectively, as odor-causing chemical substances. The detection signal data Pd shows characteristic patterns in each case, indicating that the anomaly detection system 1 according to the eleventh embodiment can detect these odor-causing chemical substances.

[0101] Furthermore, graph 13h shows an example of a radar chart display of the detection signal data Pd output by the signal processing unit 22 based on the output signal Sd of the detection unit 10 installed in the electrical equipment housing 9 when a small animal has entered the interior of the electrical equipment housing 9 and is generating the smell of small animal feces and urine or putrid odors.

[0102] The detection signal data Pd in ​​this case also shows a characteristic pattern, indicating that the abnormality detection system 1 according to the eleventh embodiment can detect the intrusion of a small animal into the electrical device casing 9. The pattern of the graph 13h appears to be a mixture of signals from odor-causing chemical substances such as skatole, indole, and ammonia.

[0103] Graphs 13a and 13b show examples of detection signal data Pd when detection unit 10 detects atmospheric gas containing the odor of a certain type of oil 1 and a type of oil 2 different from oil 1. Graphs 13c and 13d show examples of detection signal data Pd when detection unit 10 detects atmospheric gas containing the odor of a certain type of wire coating 1 and a type of wire coating 2 different from wire coating 1.

[0104] As described above, the detection signal data Pd in ​​these cases also exhibits characteristic patterns, indicating that the anomaly detection system 1 according to embodiment 11 is also capable of detecting anomalies in which the ambient gas contains these odors.

[0105] According to the eleventh embodiment, the new output signal Sd as described above can be obtained when a single odor-causing chemical substance believed to be the cause of the above-mentioned malodor or a mixture of these odors is detected as the cause of an abnormality occurring in the electrical device casing 9. Therefore, according to the eleventh embodiment, the accuracy of each abnormality detection is further improved, and it becomes possible to monitor abnormalities due to a wide range of causes. Note that the training data may be configured to use both the case where the above-mentioned component is detected alone and the case where multiple components are present, thereby improving the accuracy of discrimination.

[0106] [Software implementation example] The functional blocks of the information processing unit 20 of the anomaly detection system (particularly, the recording unit 21, the signal processing unit 22, the comparison unit 23, and the notification unit 24) may be realized by a logic circuit (hardware) formed on an integrated circuit (IC chip) or the like, or may be realized by software.

[0107] In the latter case, the anomaly detection system includes a computer that executes instructions of a program, which is software that realizes each function. This computer includes, for example, one or more processors and a computer-readable recording medium that stores the program. The object of the present invention is achieved by having the processor in the computer read and execute the program from the recording medium.

[0108] The processor may be, for example, a CPU (Central Processing Unit). The recording medium may be a "non-transitory tangible medium," such as a ROM (Read Only Memory), a tape, a disk, a card, a semiconductor memory, or a programmable logic circuit. The computer may also be provided with a RAM (Random Access Memory) for expanding the program.

[0109] The program may be supplied to the computer via any transmission medium capable of transmitting the program (such as a communication network or broadcast waves). Note that one aspect of the present invention may also be realized in the form of a data signal embedded in a carrier wave, in which the program is embodied by electronic transmission.

[0110] 〔summary〕 The abnormality detection system for electrical equipment according to aspect 1 of the present invention comprises a detection unit that monitors the atmospheric gas inside the housing of the electrical equipment, a recording unit that stores teacher data, a signal processing unit that outputs detection signal data based on the output signal of the detection unit, a comparison unit that calculates the distance or similarity between the detection signal data and the teacher data, and a notification unit that notifies of an abnormality based on the distance or similarity, wherein the detection unit is provided with two or more different odor sensors, and the teacher data and the detection signal data are data corresponding to the outputs of the plurality of odor sensors.

[0111] According to the above configuration, the influence of the relative relationship between the installation location of the detection unit and the location of the abnormality is mitigated, and an abnormality detection system for electrical equipment that can appropriately detect the occurrence of an abnormality in the electrical equipment can be realized. Furthermore, unlike conventional technology, there is no difficulty in determining a threshold value for the magnitude of the sensor output for determining an abnormality. Furthermore, since abnormalities are detected based on the sensor outputs of different odor sensors, the occurrence of false detections is reduced.

[0112] The electrical equipment anomaly detection system according to Aspect 2 of the present invention may be configured in the above-mentioned Aspect 1 such that the teacher data is generated based on a plurality of the detection signal data acquired under conditions in which a specific anomaly occurs in the electrical equipment, and the recording unit holds a plurality of the teacher data for each of the different specific anomalies. With the above-mentioned configuration, by preparing teacher data for various anomalies, it becomes possible to easily detect and distinguish between various types of anomalies in the electrical equipment.

[0113] The electrical device anomaly detection system according to Aspect 3 of the present invention may be configured in the above-mentioned Aspect 2 such that the recording unit further stores the teacher data generated based on the plurality of pieces of detection signal data acquired when the electrical device is operating normally. According to the above-mentioned configuration, by also storing the teacher data relating to the normal operating state, it becomes possible to more appropriately detect anomalies.

[0114] The electrical device anomaly detection system according to Aspect 4 of the present invention may be configured in the above-mentioned Aspect 3 so that the notification unit determines that a specific anomaly has occurred and notifies the anomaly when the distance or similarity between the detection signal data and the plurality of teacher data held by the recording unit is greatest for the specific anomaly. The above-mentioned configuration embodies a method for more appropriately detecting anomalies using teacher data related to a plurality of anomalies and teacher data related to normal operating conditions.

[0115] The electrical device anomaly detection system according to aspect 5 of the present invention is the system of aspect 1 or 2, wherein the recording unit holds the teacher data generated based on a plurality of pieces of detection signal data acquired when the electrical device is operating normally, and the alarm unit is configured to alarm an anomaly when a distance or similarity between the detection signal data and the teacher data is equal to or less than a predetermined value. With the above configuration, there is no need to acquire detection signal data and generate teacher data when an abnormality occurs in the electrical device, making it easier to build an electrical device anomaly detection system.

[0116] The electrical device abnormality detection system according to a sixth aspect of the present invention may be configured such that the detection unit is installed inside the housing in any one of the first to fifth aspects. With this configuration, it is possible to monitor the atmospheric gas inside the housing of the electrical device with a simple configuration.

[0117] The electrical device anomaly detection system according to a seventh aspect of the present invention may be configured in the sixth aspect, such that the detection unit is divided and installed in an upper part and a lower part inside the housing. This configuration increases the accuracy of anomaly detection.

[0118] The electrical device anomaly detection system according to Aspect 8 of the present invention is any one of Aspects 1 to 7, and may be configured such that the notification unit does not notify of an anomaly when the output of each of the plurality of odor sensors does not exceed a predetermined threshold value for each of the odor sensors. This configuration increases the accuracy of anomaly detection.

[0119] The electrical device abnormality detection system according to Aspect 9 of the present invention is in any one of Aspects 1 to 8, and may further include a camera that monitors the inside of the housing. According to the above configuration, the accuracy of abnormality detection is increased.

[0120] The electrical device anomaly detection system according to Aspect 10 of the present invention is any one of Aspects 1 to 9, and may further include a humidity detection unit that detects the humidity of the atmospheric gas, and the signal processing unit may be configured to output the detection signal data corrected in accordance with the humidity detected by the humidity detection unit. With the above configuration, the influence of humidity can be canceled out, and the accuracy of anomaly detection can be further improved.

[0121] An eleventh aspect of the present invention provides an electrical device anomaly detection system in any one of the first to tenth aspects, wherein the detection unit includes an odor sensor that is sensitive to at least one of odor-causing chemicals from skatole, indole, and ammonia. This configuration makes it possible to monitor anomalies caused by a wide range of factors, such as detecting the intrusion of small animals into electrical devices.

[0122] The present invention is not limited to the above-described embodiments, and various modifications are possible within the scope of the claims. Embodiments obtained by appropriately combining the technical means disclosed in different embodiments are also included in the technical scope of the present invention. [Explanation of symbols]

[0123] 1. Anomaly detection system 10. Detection unit 11, 11a, 11b Printed circuit board 12 Odor Sensor 13 Detection circuit 14a, 14b connectors 15 cases 16, 17 Suction tube 161, 171 Suction port 172 Purification filter 20 Information Processing Department 21 Recording Section 22 Signal processing section 23 Comparison section 24. Information Department 30 Humidity detection unit 9 Electrical equipment enclosures 4a, 4b, 4c, 4d Detection signal data 5a, 5b, 5c, 5d training data

Claims

1. A detection unit that monitors odors emitted by electrical equipment; a recording unit for storing teacher data; a signal processing unit that outputs detection signal data based on the output signal of the detection unit; a comparison unit that calculates a distance or similarity between the detection signal data and the teacher data; a notification unit that notifies an abnormality based on the distance or the similarity; a humidity detection unit that detects the humidity of the atmospheric gas monitored by the detection unit; Equipped with The detection unit is provided with two or more different odor sensors, the signal processing unit outputs the detection signal data corrected for each odor sensor according to the humidity detected by the humidity detection unit; The system for detecting an abnormality in an electrical device is characterized in that the teacher data and the detection signal data are data corresponding to the outputs of the plurality of odor sensors, and are data normalized for each measurement as a set including the magnitude of each output, focusing on the relative relationship between the magnitudes of the outputs of the plurality of odor sensors.

2. the teacher data is generated based on a plurality of pieces of the detection signal data acquired under a condition in which a specific abnormality occurs in the electrical equipment, 2. The electrical device abnormality detection system according to claim 1, wherein the recording unit holds a plurality of pieces of the teacher data for each of different specific abnormalities.

3. 3. The electrical equipment abnormality detection system of claim 2, wherein the recording unit further stores the teacher data generated based on a plurality of the detection signal data acquired under conditions in which the electrical equipment is operating normally.

4. 4. The electrical equipment abnormality detection system of claim 3, wherein the notification unit determines that a specific abnormality has occurred and notifies the abnormality when the similarity between the detection signal data and each of the plurality of teacher data held by the recording unit is greatest for the specific abnormality.

5. the recording unit holds the teacher data generated based on a plurality of pieces of detection signal data acquired under conditions in which the electrical device is operating normally, 3. The electrical device abnormality detection system according to claim 1, wherein the notification unit notifies of an abnormality when a similarity between the detection signal data and the teacher data is equal to or less than a predetermined value.

6. The abnormality detection system for an electrical device according to claim 1 , wherein the detection unit is installed inside a housing of the electrical device.

7. 7. The electrical device abnormality detection system according to claim 6, wherein the detection unit is divided and installed in an upper part and a lower part inside the housing.

8. 8. The electrical device abnormality detection system according to claim 1, wherein the notification unit does not notify of an abnormality when the output of each of the plurality of odor sensors does not exceed a predetermined threshold value for each of the odor sensors.

9. 8. The electrical device abnormality detection system according to claim 6, further comprising a camera for monitoring the inside of the housing.

10. 10. The electrical device abnormality detection system according to claim 1, wherein the detection unit has an odor sensor that is sensitive to at least one of odor-causing chemical substances selected from the group consisting of skatole, indole, and ammonia.

Images