Transformer condition diagnosis system, operator terminal, model learning device, and transformer condition diagnosis method

Abstract

To easily and accurately diagnose the state of a transformer.SOLUTION: A transformer state diagnostic system stores a state diagnostic model being a machine learning model that is learned with learning data constituted with information based on image data obtained by imaging oil injected into a tank of a transformer and measurement information of at least one of a substance contained in gas in the tank and a dissolved substance of the oil as an explanatory variable and with information indicating a degree of deterioration of the transformer as a target variable, and outputs the target variable obtained by inputting the explanatory variable based on the information acquired from the transformer whose state is to be diagnosed to the state diagnostic model.SELECTED DRAWING: Figure 4A

Description

[Technical Field]

[0001] The present invention relates to a transformer condition diagnosis system, an operator terminal, a model learning device, and a transformer condition diagnosis method. [Background technology]

[0002] Patent Document 1 describes a degradation diagnosis method for oil-filled transformers, which is designed to enable even an inexperienced person to accurately obtain a remaining lifespan, including the degree of degradation, as a diagnostic result in a short time. In this method, a self-organizing map is formed by learning from input data including the number of years of use of the oil-filled transformer, data quantifying the usage conditions of the oil-filled transformer, the components and concentrations of dissolved gases in the insulating oil of the oil-filled transformer, and the estimated remaining lifespan of the oil-filled transformer estimated from the components and concentrations of the dissolved gases. The input data are the components and concentrations of the dissolved gases in the insulating oil of the oil-filled transformer to be diagnosed, and the data quantifying the usage conditions of the oil-filled transformer to be diagnosed, and the estimated remaining lifespan of the oil-filled transformer to be diagnosed is obtained from the self-organizing map.

[0003] Patent Document 2 describes a transformer insulating oil degradation diagnosis system designed to perform highly accurate non-destructive testing to determine the degree of aging and abnormalities in the transformer's insulating oil. The insulating oil degradation diagnosis system includes a measurement device including a light source that irradiates the transformer's insulating oil with black light and an image capturer that captures fluorescence emitted from the object, and an analysis device that performs a degradation diagnosis of the insulating oil using information from the measurement device. The analysis device controls the operation of the light source and the image capturer, records images of the object captured by the image capturer, evaluates the recorded images of the object, and displays the evaluation results. The analysis device calculates the brightness of the image of the object, calculates the breakdown voltage of the insulating oil from the brightness, and calculates the degree of deterioration based on the calculated breakdown voltage and a preset threshold value for the insulating oil. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Japanese Patent Application Laid-Open No. 2006-229148 [Patent Document 2] Japanese Patent Application Publication No. 2018-165648 Summary of the Invention [Problem to be solved by the invention]

[0005] Oil-filled transformers deteriorate due to the effects of temperature, oxygen, moisture, etc. during long-term operation. For this reason, electric power companies and other businesses that manage transformers regularly inspect the transformer body (tank, iron core, windings, insulators, etc.), exterior parts (bushings, radiators, conservators, packing, etc.), and the oil that is poured into the tank for insulation, cooling, etc. In addition, when natural disasters such as lightning strikes or typhoons occur, workers go to the site to inspect the exterior and interior of the transformer.

[0006] When inspecting the oil at the above site, workers open the top cover of the transformer and diagnose the condition of the transformer by checking the color of the oil poured into the tank and the odor of the gas inside the tank.

[0007] However, such manual work requires sufficient experience and knowledge to accurately diagnose the condition of the transformer, so there are challenges such as ensuring diagnostic accuracy, standardizing accuracy so that it is not dependent on a specific person, and improving work efficiency.

[0008] The present invention has been made in consideration of the above background, and aims to provide a transformer condition diagnosis system, an operator terminal, a model learning device, and a transformer condition diagnosis method that are capable of easily and accurately diagnosing the condition of a transformer. [Means for solving the problem]

[0009] One of the present inventions for achieving the above object is an information processing system for diagnosing the state of a transformer (transformer state diagnosis system), which diagnoses the oil poured into the tank of the transformer. With digital still cameras Based on the captured image data Information indicating at least one of the oil color, oil transparency, and the presence or absence of sludge, obtained by The system stores a condition diagnosis model, which is a machine learning model trained using learning data configured with the explanatory variables as explanatory variables and information indicating the degree of deterioration of the transformer as a dependent variable, and outputs information based on the dependent variable obtained by inputting the explanatory variables based on the information obtained from the transformer whose condition is to be diagnosed into the condition diagnosis model.

[0010] In this way, the transformer condition diagnosis system of the present invention diagnoses the condition of a transformer using a condition diagnosis model, which is a machine learning model trained using learning data configured with image data of oil poured into the transformer tank as an explanatory variable and information indicating the degree of deterioration of the transformer as a target variable.Therefore, users such as workers on site can diagnose the condition of a transformer efficiently and with high accuracy, even if they have no knowledge or experience, simply by taking photographs of the inside of the transformer.

[0011] Another aspect of the present invention for achieving the above object is an information processing system for diagnosing the state of the transformer, which comprises: With digital still cameras Captured image data Information indicating at least one of the color of the oil, the transparency of the oil, and the presence or absence of sludge, obtained based on and information based on measurement information of at least one of the substances contained in the gas in the tank and the substances dissolved in the oil as explanatory variables, and stores a condition diagnosis model, which is a machine learning model trained using learning data configured with information indicating the degree of deterioration of the transformer as a target variable, and outputs information based on the target variable obtained by inputting the explanatory variables based on the information obtained from the transformer whose condition is to be diagnosed into the condition diagnosis model.

[0012] In this way, the transformer condition diagnosis system of the present invention diagnoses the condition of a transformer using a condition diagnosis model, which is a machine learning model trained using learning data configured with image data of oil poured into the transformer tank as an explanatory variable and information indicating the degree of deterioration of the transformer as a target variable.Therefore, a user can diagnose the condition of a transformer efficiently and with high accuracy, even if they have no knowledge or experience, by simply photographing the inside of the transformer and measuring the gas and oil present inside the transformer.

[0013] The substances contained in the gas in the tank are, for example, oxygen, nitrogen, carbon dioxide, carbon monoxide, hydrogen, methane, ethane, ethylene, acetylene, propylene, propane, and isobutane. The substances dissolved in the oil are, for example, furfural and acetone.

[0014] Another aspect of the present invention for achieving the above object is the above-mentioned transformer condition diagnosis system, which further stores an image analysis model, which is a machine learning model that uses information based on the image data as explanatory variables and determines whether the oil is reflected in the image data to an extent that diagnosis using the condition diagnosis model is possible, and if the objective variable obtained by inputting the explanatory variables based on image data taken inside the tank of the transformer whose condition is to be diagnosed into the image analysis model indicates that the oil is not reflected in the image data to an extent that diagnosis using the condition diagnosis model is possible, a message is output prompting the user to re-acquire the image data.

[0015] In this way, the transformer condition diagnosis system of the present invention uses an image analysis model to determine whether oil is reflected to an extent that diagnosis using the condition diagnosis model is possible, and if it is not reflected, it outputs a message urging the operator to re-acquire the image data. Therefore, diagnosis using the condition diagnosis model is only performed when the image data is acquired correctly, allowing the operator to work efficiently.

[0016] Another aspect of the present invention for achieving the above-mentioned object is the above-mentioned transformer condition diagnosis system, in which the contents of the objective variable obtained by inputting the explanatory variable based on the information obtained from the transformer whose condition is to be diagnosed into the condition diagnosis model are reflected in the equipment information.

[0017] In this way, the transformer condition diagnosis system automatically reflects the contents of the objective variable obtained by inputting explanatory variables based on information obtained from the transformer whose condition is to be diagnosed into the condition diagnosis model in the equipment information in which information about the transformer is managed, allowing those who manage transformers in organizations such as electric power companies to manage transformers efficiently.

[0018] Another aspect of the present invention for achieving the above-mentioned object is an information processing device (worker terminal) that constitutes the above-mentioned transformer condition diagnosis system, which is used by a worker performing work on a transformer on site, acquires the information from the transformer whose condition is to be diagnosed, stores the condition diagnosis model, and presents information based on the target variable obtained by inputting the explanatory variable based on the information acquired from the transformer whose condition is to be diagnosed into the condition diagnosis model.

[0019] In this way, the worker terminal of the present invention displays information based on the objective variables obtained by inputting explanatory variables based on information obtained from the transformer whose condition is to be diagnosed into the condition diagnosis model, so that the worker can check the diagnosis results on site and proceed with the work efficiently.

[0020] Another aspect of the present invention for achieving the above-mentioned object is an information processing device (model learning device) that constitutes the above-mentioned transformer condition diagnosis system, which generates the condition diagnosis model at any time and transmits the generated condition diagnosis model to the worker terminal of a worker performing work on the transformer on site.

[0021] In this way, the model learning device of the present invention generates learned condition diagnosis models at any time and transmits them to the worker terminals, thereby efficiently managing the condition diagnosis models of the worker terminals used by each worker to keep them up to date.

[0022] Other problems and solutions disclosed in the present application will be made clear in the detailed description and drawings. [Effects of the Invention]

[0023] According to the present invention, the state of a transformer can be diagnosed easily and accurately. [Brief explanation of the drawings]

[0024] [Figure 1] FIG. 1 is a diagram illustrating a schematic configuration of a transformer condition diagnosis system. [Figure 2] FIG. 10 is a diagram showing a worker inspecting the inside of a pole transformer on-site. [Figure 3A] This is an example of image data taken inside a transformer tank where the oil has not deteriorated. [Figure 3B] This is an example of image data taken inside a transformer tank where the oil is deteriorating. [Figure 4A] FIG. 2 is a diagram illustrating main functions of the model learning device. [Figure 4B] FIG. 2 is a diagram illustrating a hardware configuration of a model learning device. [Figure 5A] FIG. 2 is a diagram illustrating main functions of a worker terminal. [Figure 5B] FIG. 2 is a diagram illustrating a hardware configuration of a worker terminal. [Figure 6] 10 is an example of a data structure of learning data for deterioration diagnosis. [Figure 7] 10 is an example of a data structure of facility information. [Figure 8] 10 is an example of a data structure of worker terminal information. [Figure 9] 10 is an example of a data structure of a diagnosis result. [Figure 10]10 is a flowchart illustrating a learning process. [Figure 11] 10 is a flowchart illustrating a state diagnosis process. [Figure 12] 10 is a flowchart illustrating a diagnostic model receiving process. [Figure 13] 10 is a flowchart illustrating a facility information update process. DETAILED DESCRIPTION OF THE INVENTION

[0025] DETAILED DESCRIPTION OF THE INVENTION In the following description, the same or similar components may be denoted by the same reference numerals and the description thereof may be omitted.

[0026] Fig. 1 shows a schematic configuration of an information processing system (hereinafter referred to as "transformer condition diagnosis system 1") described as one embodiment of the present invention. As shown in the figure, transformer condition diagnosis system 1 includes a model learning device 100, which is an information processing device (computer) installed in a facility management department, a data center, or the like operated by an electric utility, and one or more worker terminals 200, which are information processing devices carried by workers 2 performing work on-site. The worker terminals 200 are, for example, smartphones, tablets, notebook computers, dedicated terminals, etc.

[0027] The model learning device 100 and the worker terminal 200 are communicatively connected via a communication network 5. The communication network 5 is, for example, a local area network (LAN), a wide area network (WAN), the Internet, various wireless communication networks (3G communication network, 4G communication network, 5G communication network, etc.), and various dedicated lines.

[0028] The transformer condition diagnosis system 1 supports the work performed by a worker 2 on-site when inspecting a transformer. The transformer to be inspected is an oil-filled transformer in which oil is poured into a tank for insulation and cooling purposes, such as a pole-mounted transformer, an underground transformer, or a transformer installed in a substation. The type of transformer (model, capacity, etc.) is not necessarily limited. On-site inspections of transformers are conducted periodically, and also whenever inspection becomes necessary due to natural disasters such as lightning strikes or typhoons (for example, when there is a possibility of insulation breakdown due to a lightning surge).

[0029] The transformer condition diagnosis system 1 diagnoses the condition of the transformer (e.g., degree of deterioration, presence or absence of abnormalities, whether the transformer is good or defective, etc.) by inputting, as explanatory variables, information based on image data of the inside of the transformer photographed by an operator 2 on-site and information (hereinafter referred to as "measurement information") measured on-site by various sensors (e.g., gas sensors (odor sensors), liquid concentration sensors) about the gas inside the transformer tank, into a machine learning model (hereinafter referred to as a "diagnosis model"). The system then transmits the diagnosis results to an operator terminal 200 and provides them to the user. Based on the diagnosis results, the operator 2 can efficiently diagnose the condition of the transformer on-site without relying on knowledge or experience. Furthermore, by using the diagnosis model, it is possible to detect abnormalities and symptoms that would not be noticed by manual inspection, enabling the condition of the transformer to be diagnosed with high accuracy.

[0030] The model learning device 100 learns a diagnostic model using learning data (teacher data, training data) in which explanatory variables based on image data of the interiors of a large number of transformers and measurement information acquired about each transformer are associated with objective variables (labels) received from a user (such as the degree of deterioration, the presence or absence of abnormalities, and whether the product is good or bad). The transformer used to create the learning data (hereinafter referred to as the "test transformer") may be, for example, a transformer removed from the site for replacement or inspection (a transformer that is no longer needed), or may be one specially prepared for testing or other purposes. The objective variables (labels) may be, for example, information obtained from the results of insulation measurements performed on the test transformer, measurement results using various measuring instruments, visual diagnosis results, etc.

[0031] FIG. 2 shows a situation in which an operator 2 opens the top cover (not shown) of a transformer 3 (a pole-mounted transformer in this example) at the site and uses an operator terminal 200 to inspect the condition inside the tank 31 of the transformer 3. The operator 2 photographs the inside of the tank 31 so that the oil 32 poured into the tank 31 is properly captured. For example, a camera provided in the operator terminal 200 is used as the photographing equipment. For example, a polarizing filter (PL filter) may be used to suppress light reflection from the oil surface so that the condition of the oil 32 is clearly captured. Furthermore, a mechanism may be provided that uses, for example, an augmented reality (AR) mechanism to guide the operator 2 in the photographing direction and angle of view so that the photographing direction and angle of view are in a state necessary to ensure the diagnostic accuracy of the diagnostic model.

[0032] 3A and 3B are both examples of image data photographed showing the state inside the tank 31 of the transformer 3. Of these, Fig. 3A is image data photographed showing the state inside the tank 31 of the transformer where the oil 32 has not deteriorated, and Fig. 3B is image data photographed showing the state inside the tank 31 of the transformer 3 where the oil 32 has deteriorated more than in the state shown in Fig. 3A.

[0033] As shown in these figures, as the deterioration of oil 32 progresses, the color and transparency of oil 32 change (from pale yellow to brown), eventually resulting in the formation of sludge (a compound of oil 32 with metal and coil insulation as the acid value increases). When sludge adheres to the coil insulation, iron core, and heat dissipation surface, the cooling effect decreases, the temperature rises significantly, and the thermal deterioration of the insulation accelerates. Furthermore, if operation continues in a state of insulation deterioration, partial discharge occurs due to overvoltage, etc., and electrical / mechanical stress during external surges and external short circuits leads to insulation breakdown. Furthermore, the oil 32 itself becomes more water-absorbent as the degradation products dissolve, reducing its insulating properties.

[0034] 4A shows the main functions of the model learning device 100. As shown in the figure, the model learning device 100 has the following functions: a storage unit 110, an information acquisition management unit 120, a learning data generation unit 130, a diagnostic model learning unit 135, a diagnostic model distribution unit 140, a diagnostic result receiving unit 145, and an equipment information update unit 150.

[0035] Of the above functions, the storage unit 110 stores image data 111, measurement information 112, learning data 113, diagnostic model 114, equipment information 115, and worker terminal information 116.

[0036] Of these, the image data 111 is image data of a predetermined data type, which is obtained by photographing the state inside the tank 31 of the transformer 3. The image data 111 includes information indicating the date and time when the image data 111 was photographed. The image data 111 may be data (frame data) extracted from a moving image.

[0037] The measurement information 112 is measurement information (concentration of components, etc.) measured by various sensors (gas sensors, liquid concentration sensors, etc.) regarding substances contained in the gas present in the tank 31 of the transformer 3 and dissolved substances in the oil.

[0038] The learning data 113 is used for learning the diagnostic model 114 (an image analysis model 1141 and a condition diagnostic model 1142 described later), and includes information based on the image analysis learning data 1131 and the degradation diagnosis learning data 1132.

[0039] Of these, the learning data for image analysis 1131 is data in which image data of the inside of the tank 31 of the transformer 3 is used as an explanatory variable, and the explanatory variable is associated with a target variable (information indicating whether or not oil 32 is reflected in the image data, and information indicating the area in which oil 32 is reflected in the image data).

[0040] In addition, the learning data 1132 for deterioration diagnosis is data in which image data of the inside of the tank 31 of the transformer 3 and measurement information measured on the components of various substances contained in the gas and oil present in the tank 31 are used as explanatory variables, and the explanatory variables are associated with target variables (the degree of deterioration of the transformer 3, whether there is an abnormality in the transformer 3, whether the transformer 3 is good or defective, etc.).

[0041] Furthermore, the learning data for deterioration diagnosis 1132 only needs to include at least image data as explanatory variables, and whether or not measurement information is used as an explanatory variable, and if so, what type of measurement information to use, may be determined according to the need and the situation.

[0042] The diagnostic model 114 is a machine learning model that is trained using the training data 113 (image analysis training data 1131, deterioration diagnosis training data 1132). The diagnostic model 114 includes an image analysis model 1141, which is a machine learning model that analyzes image data, and a condition diagnosis model 1142, which is a machine learning model that diagnoses the condition (degree of deterioration) of the oil 32 based on the image data. The entity of the diagnostic model 114 is, for example, a polynomial, a determinant, a mathematical formula, a vector, or the like, including adjustable parameters. The type of the diagnostic model 114 is not necessarily limited, but may be realized, for example, by deep learning (DNN (Deep Neural Network), CNN (Convolutional Neural Network), or the like).

[0043] Of the diagnostic models 114, the image analysis model 1141 determines whether or not the image data shows enough oil 32 to enable diagnosis (analysis) by the condition diagnostic model 1142, and identifies the area in the image data where the oil 32 is shown. When an explanatory variable (hereinafter referred to as "target transformer information") acquired for the transformer 3 to be diagnosed (hereinafter referred to as "target transformer") is input, the condition diagnostic model 1142 outputs the above-mentioned objective variable.

[0044] The image analysis model 1141 learns, for example, feature quantities (color, shade, pattern, etc.) of images in which the oil 32 is present and images in which the oil is not present. The image analysis model 1141 also learns, for example, feature quantities (shape, color, etc.) of an area in which the oil 32 is present and an area in which other components of the transformer 3 are present. The image analysis model 1141 identifies the oil 32 and the area in which the oil 32 is present (detects objects) in the image data using, for example, a sliding window method, a histograms of oriented gradients (HOG) feature quantity, a region proposal method, a Faster R-CNN, a YOLO (You Only Look Once), a single shot detector (SSD), or end-to-end learning. The image analysis model 1141 may be, for example, a machine learning model (anomaly detection model) trained by an autoencoder (such as a variational autoencoder (VAE)) using image data in which the oil 32 is not present as a steady state.

[0045] The features used in the diagnostic model 114 (image analysis model 1141, condition diagnostic model 1142) may be set manually by a human system, or may be automatically extracted by an information processing device using, for example, a known feature extraction method (SIFT (Scale-Invariant Feature Transform), SURF (Speeded-Up Robust Features), FAST (Features from Accelerated Segment Test), BRIEF (Binary Robust Independent Elementary Features), ORB (Oriented FAST and Rotated BRIEF), etc.).

[0046] The equipment information 115 is information used by electric utilities and the like to manage the transformers 3, and includes information about the transformers 3 installed in the area (installation location, model, capacity, installation date and time, past inspection dates and inspection results, diagnosis results by the diagnostic model 114, etc.).

[0047] The worker terminal information 116 includes information about the worker terminal 200 to which the diagnostic model 114 is delivered from the model learning device 100.

[0048] Among the functions shown in the figure, an information acquisition management unit 120 acquires image data and measurement information, and manages these pieces of information as image data 111 and measurement information 112, respectively.

[0049] The learning data generation unit 130 generates the learning data 113. For example, the learning data generation unit 130 generates the learning data 113 by presenting explanatory variables (information based on image data or measurement information) and receiving objective variables to be set for the explanatory variables from the user via a user interface.

[0050] The diagnostic model learning unit 135 uses the learning data 113 to learn the diagnostic model 114 .

[0051] The diagnostic model distribution unit 140 distributes (transmits) the latest diagnostic model 114 learned by the diagnostic model learning unit 135 to the operator terminal 200 via the communication network 5.

[0052] The diagnosis result receiving unit 145 receives the diagnosis result sent from the worker terminal 200 via the communication network 5 .

[0053] The equipment information update unit 150 receives the diagnosis results sent from the operator terminal 200 and updates the equipment information 115 with the contents of the received diagnosis results.

[0054] 4B is a diagram showing the hardware configuration of model learning device 100. As shown in the figure, model learning device 100 includes processor 11, main memory device 12, auxiliary memory device 13, input device 14, output device 15, and communication device 16. Model learning device 100 may be configured from multiple information processing devices connected to each other so that they can communicate with each other.

[0055] All or part of the configuration shown in the figure may be realized using virtual information processing resources provided using virtualization technology, process space separation technology, or the like, such as a virtual server provided by a cloud system. All or part of the functions of model learning device 100 may be realized, for example, by a service provided by a cloud system via an API (Application Programming Interface). All or part of the functions of model learning device 100 may be realized, for example, using Software as a Service (SaaS), Platform as a Service (PaaS), Infrastructure as a Service (IaaS), or the like.

[0056] In the same figure, the processor 11 is configured using, for example, a CPU (Central Processing Unit), an MPU (Micro Processing Unit), a GPU (Graphics Processing Unit), an FPGA (Field Programmable Gate Array), an ASIC (Application Specific Integrated Circuit), an AI (Artificial Intelligence) chip, etc.

[0057] Main memory device 12 is a device used by processor 11 when executing a program, and is, for example, a read-only memory (ROM), a random access memory (RAM), or a non-volatile memory (NVRAM (Non-Volatile RAM)). The various functions of model learning device 100 are realized by processor 11 reading programs and data stored in auxiliary memory device 13 into main memory device 12 and executing them.

[0058] The auxiliary storage device 13 is, for example, an SSD (Solid State Drive), a hard disk drive, an optical storage device (CD (Compact Disc), DVD (Digital Versatile Disc), etc.), a storage system, a reading / writing device for a recording medium such as an IC card, an SD card, or an optical recording medium, a storage area of ​​a cloud server, etc. Programs and data can be read into the auxiliary storage device 13 via a recording medium reading device or a communication device 16. The programs and data stored (memorized) in the auxiliary storage device 13 are read into the main storage device 12 as needed.

[0059] The input device 14 is an interface that accepts input from the outside, and is, for example, a touch panel, a keyboard, a mouse, a card reader, a pen-input tablet, a voice input device, or the like.

[0060] Output device 15 is an interface that outputs various information such as processing progress and processing results. Output device 15 is, for example, a display device (liquid crystal monitor, LCD (Liquid Crystal Display), graphic card, etc.) that visualizes the various information described above, a device that converts the various information described above into audio (audio output device (speaker, etc.)), or a device that converts the various information described above into text (printer, etc.). Note that, for example, model learning device 100 may be configured to input and output information to and from other devices via communication device 16.

[0061] The input device 14 and the output device 15 constitute a user interface that receives information from the user and presents information to the user.

[0062] The communication device 16 is a device that realizes communication with other devices. The communication device 16 is a wired or wireless communication interface that realizes communication with other devices via a communication network 5 such as the Internet, and is, for example, a NIC (Network Interface Card), a wireless communication module, a USB module, or the like.

[0063] The model learning device 100 may be equipped with, for example, an operating system, a file system, a DBMS (DataBase Management System) (relational database, NoSQL, etc.), a KVS (Key-Value Store), etc.

[0064] Each of the above-described functions of model learning device 100 is realized by processor 11 reading and executing a program stored in main memory device 12, or by hardware (FPGA, ASIC, AI chip, etc.) that constitutes model learning device 100. Model learning device 100 stores the above-described various pieces of information (data), for example, as a database table or a file managed by a file system.

[0065] 5A shows the main functions of the worker terminal 200. As shown in the figure, the worker terminal 200 has the following functions: a storage unit 210, a target transformer information acquisition unit 220, a transformer state diagnosis unit 225, a diagnosis result presentation unit 230, a diagnosis result transmission unit 235, and a diagnostic model update unit 240.

[0066] Of the above functions, the storage unit 210 stores target transformer information 211, the diagnostic model 114, and the diagnostic result 213.

[0067] Of these, the target transformer information 211 is information (image data, measurement information) used as explanatory variables to be input to the diagnostic model 114 when the transformer state diagnostic unit 225 diagnoses the state of the target transformer.

[0068] The diagnostic model 114 is distributed from the model learning device 100. The diagnostic model 114 is the same as that described above, so a description thereof will be omitted.

[0069] The diagnosis result 213 includes information about the diagnosis result that the transformer state diagnosis unit 225 performed on the target transformer using the diagnosis model 114 .

[0070] Of the above functions, the target transformer information acquisition unit 220 acquires information used as explanatory variables from the imaging device 27 and various sensors 28 described below, and manages the acquired information as target transformer information 211.

[0071] The transformer state diagnosis unit 225 inputs the target transformer information 211 to the diagnosis model 114 as an explanatory variable, and thereby obtains the objective variable output by the diagnosis model 114.

[0072] The diagnosis result presentation unit 230 presents information based on the objective variables acquired by the transformer state diagnosis unit 225 to the user as a diagnosis result, and manages the diagnosis result as the diagnosis result 213 .

[0073] The diagnostic result transmission unit 235 transmits the diagnostic result 213 to the model learning device 100 via the communication network 5.

[0074] The diagnostic model update unit 240 receives the diagnostic model 114 sent from the model learning device 100 via the communication network 5, and updates the diagnostic model 114 stored in the storage unit 210 with the received diagnostic model 114. Note that the diagnostic model update unit 240 may receive only the update difference of the diagnostic model from the model learning device 100, and update the diagnostic model 114 with the received update difference.

[0075] 5B shows an example of the hardware configuration of the worker terminal 200. As shown in the figure, the worker terminal 200 includes a processor 21, a main memory device 22, an auxiliary memory device 23, an input device 24, an output device 25, a communication device 26, an image capture device 27, and various sensors 28. Of these, the processor 21, the main memory device 22, the auxiliary memory device 23, the input device 24, the output device 25, and the communication device 26 are similar to the processor 11, the main memory device 12, the auxiliary memory device 13, the input device 14, the output device 15, and the communication device 16 of the model learning device 100, respectively, and therefore description thereof will be omitted.

[0076] The photographing device 27 is a device that generates image data by photographing a subject. If the worker terminal 200 is, for example, a smartphone, a tablet, or a notebook computer, the photographing device 27 is, for example, a digital still camera provided therein.

[0077] The various sensors 28 are one or more sensors that detect the concentrations of various substances contained in the gas present in the tank 31 of the transformer 3. The various sensors 28 are, for example, gas sensors (odor sensors) that detect substances such as oxygen, nitrogen, carbon dioxide, carbon monoxide, hydrogen, methane, ethane, ethylene, acetylene, propylene, propane, and isobutane. The various sensors 28 are, for example, liquid concentration sensors such as infrared liquid component concentration meters. Substances that are detected by the liquid concentration sensors are, for example, dissolved substances (furfural, acetone) in the oil 32 that correlate with the degree of deterioration of the insulating paper.

[0078] 6 shows an example of the data structure of the degradation diagnosis learning data 1132. As shown in the figure, the illustrated degradation diagnosis learning data 1132 is made up of one or more records each having items such as a data ID 611, image data 612, measurement information 613, degree of degradation 614, and presence or absence of foreign matter 615. One record of the degradation diagnosis learning data 1132 corresponds to one piece of degradation diagnosis learning data.

[0079] Of the above items, image data 612 and measurement information 613 (or feature quantities extracted from these pieces of information) correspond to explanatory variables. Furthermore, the degree of degradation 614 and the presence or absence of foreign matter 615 correspond to objective variables (labels). For example, the user sets the contents of the objective variables corresponding to the explanatory variables while checking the contents of the explanatory variables displayed on the screen via a user interface provided by the model learning device 100.

[0080] 7 shows an example of the data structure of the equipment information 115. As shown in the figure, the illustrated equipment information 115 is made up of one or more records each having items such as a transformer ID 711, model 712, capacity 713, installation location 714, installation date and time 715, last diagnosis date 716, and diagnosis result 717. One record of the equipment information 115 corresponds to one transformer 3.

[0081] Of the above items, the transformer ID 711 stores the identifier of the transformer 3 (hereinafter referred to as the "transformer ID"). The model 712 stores information indicating the model of the transformer 3. The capacity 713 stores information indicating the capacity of the transformer 3. The installation date and time 715 stores the date and time when the transformer 3 was installed. The last diagnosis date 716 is set to the date on which the most recent diagnosis was performed on the transformer 3. The diagnosis result 717 stores the diagnosis result of the transformer 3 (the diagnosis result received from the operator terminal 200).

[0082] 8 shows an example of the data structure of worker terminal information 116. As shown in the figure, the illustrated worker terminal information 116 is made up of one or more records each having items such as a terminal ID 811, a network address 812, a current version 813, and a last update date and time 814. One record of the worker terminal information 116 corresponds to one worker terminal 200.

[0083] The terminal ID 811 stores an identifier of the worker terminal 200 (hereinafter referred to as "terminal ID"). The NW address 812 stores a network address (such as an IP address) assigned to the worker terminal 200. The current version 813 stores information indicating the current version of the diagnostic model 114 stored in the worker terminal 200 (the current versions of the image analysis model 1141 and the condition diagnosis model 1142). The last update date and time 814 stores the date and time of the most recent update of the diagnostic model 114 of the worker terminal 200.

[0084] 9 shows an example of the data structure of the diagnostic results 213. As shown in the figure, the illustrated diagnostic results 213 include the contents of the following items: transformer ID 911, diagnostic date and time 912, deterioration level 913, and presence or absence of foreign matter 914. One diagnostic result 213 corresponds to the result of one diagnosis performed on one transformer 3.

[0085] The transformer ID 911 stores the transformer ID. The diagnosis date and time 912 stores the date and time when the diagnosis for the diagnosis result was performed. The deterioration level 913 stores the deterioration level diagnosed by the diagnosis model 114. The presence or absence of foreign matter 914 stores information indicating the presence or absence of foreign matter in the oil 32 in the tank 31 of the transformer 3, diagnosed by the diagnosis model 114.

[0086] Next, the main processes performed in the transformer condition diagnosis system 1 will be described.

[0087] 10 is a flowchart illustrating the process (hereinafter referred to as "learning process S1000") that the diagnostic model learning unit 135 of the model learning device 100 performs when learning the diagnostic model 114. The learning process S900 will be described below with reference to this figure.

[0088] First, the diagnostic model learning unit 135 reads the image analysis learning data 1131 and the degradation diagnosis learning data 1132 from the auxiliary storage device 13 to the main storage device 12 (S1011).

[0089] Next, the diagnostic model learning unit 135 uses the image analysis learning data 1131 to learn the image analysis model 1141 (S1012).

[0090] Next, the diagnostic model learning unit 135 uses the degradation diagnosis learning data 1132 to learn the state diagnostic model 1142 (S1013).

[0091] Next, the diagnostic model distribution unit 140 distributes (transmits) the latest trained diagnostic model 114 (image analysis model 1141, condition diagnosis model 1142) to each worker terminal 200 via the communication network 5 (S1014).

[0092] Since the learning of the diagnostic model 114 requires a large amount of information processing resources, the learning may be performed using, for example, an API (Application Programming Interface) for machine learning provided by a cloud system.

[0093] Furthermore, the diagnostic model learning unit 135 may, for example, verify the prediction accuracy of the trained diagnostic model 114. In this case, for example, the learning data is classified in advance into data for learning and data for verification, and the learning data is used for learning, and the verification data is used for verification.

[0094] FIG. 11 is a flowchart illustrating the main process performed by the worker terminal 200 (hereinafter referred to as "condition diagnosis process S1100"). The worker 2 brings the worker terminal 200 to the site and uses the worker terminal 200 to inspect the transformer 3. The condition diagnosis process S1100 is the main process performed by the worker terminal 200 at that time. The condition diagnosis process S1100 may be performed independently by the worker terminal 200, or may be performed by the worker terminal 200 in cooperation with a cloud service via the communication network 5. The condition diagnosis process S1100 is started, for example, when the worker 2 performs a predetermined start operation on the worker terminal 200. The condition diagnosis process S1100 will be described below with reference to FIG.

[0095] First, the target transformer information acquisition unit 220 of the worker terminal 200 acquires image data captured by the worker 2 operating the photographing device 27 (S1111). At this time, the target transformer information acquisition unit 220 may present information about the target transformer (for example, information acquired from the equipment information 115 of the model learning device 100, etc.) to the worker 2.

[0096] Next, the transformer state diagnosis unit 225 of the operator terminal 200 inputs the acquired image data into the image analysis model 1141 to determine whether or not the image data shows oil 32 (whether or not the image data shows oil 32 in a state that allows it to be used to diagnose the degree of deterioration) (S1112). If the transformer state diagnosis unit 225 determines that the image data does not show oil 32 (S1112: NO), the process returns to S1111. In this case, the operator terminal 200 may display a message urging the operator 2 to take a picture again (reacquire image data). If the transformer state diagnosis unit 225 determines that the image data shows oil 32 (S1112: YES), the process proceeds to S1113.

[0097] In S1113, the target transformer information acquisition unit 220 acquires measurement information from the various sensors 28.

[0098] Next, the transformer state diagnosis unit 225 inputs the image data acquired in S1111 and the measurement information acquired in S1113 to the state diagnosis model 1142, and acquires the objective variable output by the state diagnosis model 1142 (S1114).

[0099] Next, the diagnosis result presenting unit 230 displays the diagnosis result based on the acquired objective variables (S1115).

[0100] Next, the diagnosis result transmitting unit 235 transmits the diagnosis result to the model learning device 100 via the communication network 5 (S1116).

[0101] In this way, the worker 2 can accurately diagnose the degree of deterioration of the target transformer even without knowledge or experience, simply by taking photographs of the inside of the transformer 3 and measuring the gas and oil 32 present inside the tank 31 of the target transformer. This allows the inspection work of the transformer 3 to be carried out efficiently and reliably, significantly reducing the workload of the worker 2.

[0102] 12 is a flowchart illustrating the process (hereinafter referred to as "diagnostic model reception process S1200") that the operator terminal 200 performs when updating the diagnostic model 114. The diagnostic model reception process S1200 will be described below with reference to this figure.

[0103] The diagnostic model update unit 240 of the operator terminal 200 constantly monitors whether a new diagnostic model 114 (the latest diagnostic model 114 learned using the latest learning data 113) or an update difference has been received from the model learning device 100 (S1211). The diagnostic model distribution unit 140 of the model learning device 100 distributes the diagnostic model to the operator terminal 200 at an appropriate timing, such as when the diagnostic model 114 has been updated with new learning data 113.

[0104] When the diagnostic model update unit 240 receives a new diagnostic model 114 or an update difference from the model learning device 100, it updates the diagnostic model 114 to the content of the received diagnostic model 114 (or reflects the update difference in the diagnostic model 114) (S1212).

[0105] In this way, the diagnostic model 114 of the operator terminal 200 is always kept up to date by learning using the latest learning data 113, so that the deterioration level of the transformer 3 can be diagnosed accurately, stably and continuously on-site.

[0106] 13 is a flowchart illustrating the process (hereinafter referred to as "facility information update process S1300") performed by the model learning device 100 when updating the facility information 115. The facility information update process S1300 will be described below with reference to this figure.

[0107] The equipment information update unit 150 of the model learning device 100 constantly monitors whether or not a diagnosis result has been received from the operator terminal 200 (S1311).

[0108] When the equipment information update unit 150 receives a new diagnosis result from the operator terminal 200, it updates the received diagnosis result to the information of the corresponding transformer 3 in the equipment information 115 (S1312).

[0109] In this way, the diagnosis results of the transformers 3 installed at various locations are managed in a unified manner as the facility information 115, so that, for example, a person in charge of managing the transformers 3 in an organization such as an electric utility can efficiently manage the transformers 3. The diagnosis results received from the operator terminal 200 may be used, for example, as a target variable (label) when generating the learning data 113.

[0110] Although the embodiments of the present invention have been described in detail above, the above description is intended to facilitate understanding of the present invention and is not intended to limit the present invention. The present invention may be modified or improved without departing from the spirit thereof, and the present invention naturally includes equivalents thereof. For example, the above embodiments have been described in detail to clearly explain the present invention, and the present invention is not necessarily limited to those including all of the described configurations. Furthermore, it is possible to add, delete, or replace part of the configuration of the above embodiments with other configurations. [Explanation of symbols]

[0111] 1 Transformer condition diagnosis system, 2 Worker, 3 Transformer, 27 Photography device, 28 Various sensors, 31 Tank, 32 Oil, 5 Communication network, 100 Model learning device, 110 Memory unit, 111 Image data, 112 Measurement information, 113 Learning data, 1131 Learning data for image analysis, 1132 Learning data for deterioration diagnosis, 114 Diagnostic model, 1141 Image analysis model, 1142 Condition diagnosis model, 115 Equipment information, 116 Worker terminal information, 120 Information acquisition management unit, 130 Learning data generation unit, 135 Diagnostic model learning unit, 140 Diagnostic model distribution unit, 145 Diagnostic result receiving unit, 150 Equipment information update unit, 200 Worker terminal, 210 Memory unit, 211 Target transformer information, 213 Diagnostic result, 220 Target transformer information acquisition unit, 225 Transformer condition diagnosis unit, 230 diagnosis result presentation unit, 235 diagnosis result transmission unit, 240 diagnosis model update unit, S1000 learning process, S1100 condition diagnosis process, S1200 diagnosis model reception process, S1300 equipment information update process

Claims

1. An information processing system for diagnosing a condition of a transformer, comprising: A condition diagnosis model is stored, which is a machine learning model trained using learning data configured using information indicating at least one of the color of the oil, the transparency of the oil, and the presence or absence of sludge, which is acquired based on image data obtained by photographing the oil poured into the tank of the transformer with a digital still camera, as an explanatory variable, and information indicating the degree of deterioration of the transformer as a target variable; outputting information based on a target variable obtained by inputting the explanatory variables based on the information acquired from the transformer whose state is to be diagnosed into the state diagnosis model; Transformer condition diagnostic system.

2. An information processing system for diagnosing a condition of a transformer, comprising: a condition diagnosis model that is a machine learning model trained using learning data configured using explanatory variables including information indicating at least one of the color of the oil, the transparency of the oil, and the presence or absence of sludge, which is obtained based on image data of the oil poured into the tank of the transformer photographed by a digital still camera, and information based on measurement information of at least one of substances contained in the gas in the tank and substances dissolved in the oil, and information indicating the degree of deterioration of the transformer as a response variable; outputting information based on a target variable obtained by inputting the explanatory variables based on the information acquired from the transformer whose state is to be diagnosed into the state diagnosis model; Transformer condition diagnostic system.

3. 3. The transformer condition diagnosis system according to claim 1 or 2, Further storing an image analysis model, which is a machine learning model that has learned feature amounts of image data when oil is reflected and when oil is not reflected, and that uses information based on the image data as explanatory variables to determine whether the oil is reflected in the image data to an extent that diagnosis by the condition diagnosis model is possible; outputting a message urging the user to reacquire the image data when a response variable obtained by inputting the explanatory variables based on image data of the inside of the tank of the transformer whose condition is to be diagnosed into the image analysis model indicates that the oil is not reflected in the image data to an extent that allows diagnosis by the condition diagnosis model; Transformer condition diagnostic system.

4. 3. The transformer condition diagnosis system according to claim 1 or 2, storing facility information relating to each of the plurality of transformers to be managed; The explanatory variables based on the information acquired from the transformer whose state is to be diagnosed are input into the state diagnosis model, and the contents of the objective variables are reflected in the equipment information. Transformer condition diagnostic system.

5. 3. The transformer condition diagnosis system according to claim 1 or 2, The oil is a mineral oil-based insulating oil. Transformer condition diagnostic system.

6. 3. The transformer condition diagnostic system according to claim 2, The substance contained in the gas in the tank is at least one of oxygen, nitrogen, carbon dioxide, carbon monoxide, hydrogen, methane, ethane, ethylene, acetylene, propylene, propane, and isobutane. Transformer condition diagnostic system.

7. 3. The transformer condition diagnostic system according to claim 2, The dissolved substance of the oil is at least one of furfural and acetone. Transformer condition diagnostic system.

8. An information processing device constituting the transformer condition diagnosis system according to claim 1, Used by workers performing on-site work on transformers, obtaining said information from the transformer whose condition is to be diagnosed; storing the condition diagnosis model; presenting information based on the objective variable obtained by inputting the explanatory variables based on the information acquired from the transformer whose state is to be diagnosed into the state diagnosis model; Worker terminal.

9. An information processing device constituting the transformer condition diagnosis system according to claim 1, generating the condition diagnosis model as needed, transmitting the generated state diagnosis model to an operator terminal of an operator performing on-site work on the transformer; Model learning device.

10. 1. A method for diagnosing a condition of a transformer, comprising: An information processing device having a processor and a memory, a step of storing a condition diagnosis model, which is a machine learning model trained using learning data configured using information indicating at least one of the color of oil, the transparency of oil, and the presence or absence of sludge, which is acquired based on image data of oil poured into a tank of a transformer photographed by a digital still camera, as an explanatory variable, and information indicating the deterioration level of the transformer as a target variable; and a step of outputting information based on a response variable obtained by inputting the explanatory variables based on the information acquired from the transformer whose state is to be diagnosed into the state diagnosis model; A transformer condition diagnosis method is provided.

11. 1. A method for diagnosing a condition of a transformer, comprising: An information processing device having a processor and a memory, a step of storing a condition diagnosis model, which is a machine learning model trained using learning data configured using information indicating the degree of deterioration of the transformer as a response variable, and information indicating at least one of the color of the oil, the transparency of the oil, and the presence or absence of sludge, which is obtained based on image data of the oil poured into the tank of the transformer photographed by a digital still camera, and information based on measurement information of at least one of substances contained in the gas in the tank and substances dissolved in the oil, as explanatory variables; and a step of outputting information based on a response variable obtained by inputting the explanatory variables based on the information acquired from the transformer whose state is to be diagnosed into the state diagnosis model; A transformer condition diagnosis method is provided.

Images