Fault monitoring device and method for multimode data fusion

Abstract

The invention provides a multi-mode data fusion fault monitoring device and method, and the device comprises a sensing terminal and a holder, the bottom of the sensing terminal is provided with the holder through a holder quick-mounting plate, one side of the holder is provided with a switching power supply, and the sensing terminal is in communication connection with the holder through an RS485 communication line. The beneficial effects of the invention are that the system can integrate infrared, image, sound and partial discharge detection functions, and achieves the multi-dimensional monitoring of power transformation equipment. According to the invention, a machine learning algorithm and an intelligent diagnosis technology are adopted, and the accuracy and sensitivity of fault identification are improved. By means of the pan-tilt design, the device has the omni-directional monitoring capacity, and monitoring blind areas are reduced. According to the distributed installation mode, the expandability and deployment flexibility of the device are improved. According to the invention, real-time early warning is realized, and the response time of equipment faults can be effectively shortened.

Description

Technical Field

[0001] The invention belongs to the field of power fault monitoring, and in particular relates to a multi-mode data fusion fault monitoring device and method. Background Art

[0002] With the rapid development of power systems, the safe and stable operation of substation equipment is particularly important. However, due to the complex operating environment of substation equipment, equipment aging or sudden failures may lead to serious economic losses and power supply safety issues. At present, traditional equipment monitoring methods are usually single and cannot effectively integrate multiple detection data, making it difficult to detect and locate equipment failures in a timely manner. Therefore, there is an urgent need for an intelligent device that can integrate multiple sensing methods and has real-time diagnostic capabilities. Summary of the invention

[0003] In view of this, the present invention aims to propose a multi-mode data fusion fault monitoring device and method to solve at least one problem existing in the above-mentioned prior art. The device and method can integrate multiple monitoring functions such as infrared, image and partial discharge (PD), sound, temperature, humidity, light sensing sensors, etc., and can be used in combination with pan-tilt components and installed in a distributed manner near substation equipment to achieve real-time monitoring and diagnosis of key equipment in substations.

[0004] To achieve the above object, the technical solution of the present invention is achieved as follows: In a first aspect, the present invention provides a multi-mode data fusion fault monitoring device, comprising a sensing terminal and a pan / tilt, wherein the pan / tilt is installed at the bottom of the sensing terminal through a pan / tilt quick-release plate, a switching power supply is installed on one side of the pan / tilt, and the sensing terminal is connected to the pan / tilt through an RS485 communication line; The hardware structure of the perception terminal includes a main control module, a power module, a visible light image acquisition module, an infrared acquisition module, a partial discharge acquisition module, a sound acquisition module, a temperature and humidity acquisition module, a data transmission module, a security module, a data storage module and an input and output module, and the power module, the visible light image acquisition module, the infrared acquisition module, the partial discharge acquisition module, the sound acquisition module, the temperature and humidity acquisition module, the data transmission module, the security module, the data storage module and the input and output module are all communicatively connected with the main control module; The pan-tilt head comprises a switching power supply and a pan-tilt head module. The switching power supply is used to supply power to the pan-tilt head module, and the pan-tilt head module is communicatively connected with the main control module.

[0005] Furthermore, the main control module uses an ARM A7 quad-core processor to implement the scheduling of each module; The power module includes a high-voltage terminal, a 220VAC to 5V module, a DCDC module and a POE module. The high-voltage terminal is connected to the DCDC module through the 220VAC to 5V module, the POE module is also connected to the DCDC module, and the DCDC module is connected to the main control module; The visible light image acquisition module includes a visible light camera, an infrared fill light, a light sensing sensor and an infrared filter, all of which are connected to the main control module for communication; The infrared acquisition module includes an infrared sensor, which is used to collect temperature information of the substation equipment and generate a thermal image; The partial discharge acquisition module includes a partial discharge antenna, a partial discharge sensor and a conditioning circuit. The input end of the conditioning circuit is connected to the main control module for communication. The conditioning circuit is connected to the partial discharge antenna through the partial discharge sensor. The sound collection module includes a sound sensor that is communicatively connected to the main control module; The temperature and humidity acquisition module includes a temperature and humidity sensor that is communicatively connected to the main control module; The data transmission module includes a wired communication module and a wireless communication module. The wired communication module is mainly composed of RJ45, a network card chip, etc. The wireless communication module includes a 4G communication module, a LoRa communication module, and a WiFi / BT / GPS three-in-one module. The security module includes an encryption chip that is communicatively connected to the main control module; The data storage module includes a TF card that is communicatively connected to the main control module; The input-output module includes an indicator light that is communicatively connected to the main control module.

[0006] Furthermore, the software structure of the perception terminal includes a data acquisition module, a data management module, a security module, a data transmission module, an input / output module, a power module and a convergence module. The data management module is communicated with the data acquisition module and the security module respectively. The security module is also communicated with the data transmission module and the main control module respectively. The input / output module, the power module and the convergence module are all communicated with the main control module.

[0007] Furthermore, the visible light image acquisition module, infrared acquisition module, partial discharge acquisition module, sound acquisition module, and temperature and humidity acquisition module constitute a data acquisition module, which is responsible for collecting sensor data and dispatching GPS and PTZ to complete positioning and rotation of monitoring angles; The data management module is responsible for the preprocessing, analysis and diagnosis of sampled data, as well as the storage of related data and files; The security module is responsible for data encryption and decryption to ensure secure data transmission; The data transmission module is responsible for data encapsulation and analysis, and completes the sending and receiving of data and instructions through wired and wireless means; The input-output module is used for user interaction of the device; The power supply module is used to ensure the power supply and power consumption control of the system; The aggregation module expands and implements the data access function of other sensor devices through LoRa.

[0008] In a second aspect, based on the same concept, the present invention also provides a multi-mode data fusion fault monitoring method, comprising the following steps: S1, data collection based on data collection module; S2, performing data preprocessing on the data processed in step S1 based on the data management module; S3, analyzing and diagnosing the data processed in step S2 based on the data management module; S4, storing the data processed in step S3 based on the data management module; S5. Encrypt and decrypt the data processed in step S4 based on the security module; S6. Push an alarm for the data processed in step S5 based on the data transmission module.

[0009] Furthermore, in step S1, data collection includes: S11, sampling scheduling, and judging whether sampling is required, if yes, proceed to step S12 and step S13 at the same time, if no, continue sampling scheduling; S12, controlling the pan / tilt component to rotate according to the preset position, and proceeding to step S14; S13, collecting the latitude and longitude and location information of the monitoring device; S14, start multi-channel sampling of sensors; The sensor's multi-channel sampling includes visible light collection, partial discharge collection, infrared collection, sound collection, and temperature and humidity collection.

[0010] Furthermore, in step S14, the visible light collection includes: First, the light sensor is used to analyze the brightness of the light around the device, and the infrared filter and infrared fill light are linked to assist in the collection of visible light; When the light sensor detects a dark scene, first turn on IRCUT, then turn on the infrared fill light, and finally call the visible light camera to take pictures.

[0011] Further, in step S2, data preprocessing includes visible light image preprocessing, partial discharge data preprocessing, infrared data preprocessing and sound data preprocessing; Among them, visible light image preprocessing includes distortion processing and brightness optimization; Partial discharge data preprocessing includes eigenvalue extraction and feature map drawing; Infrared data preprocessing includes filtering and compensation processing of temperature frame data, extraction of temperature extreme value for zone monitoring, and generation of infrared thermal imaging images; Sound data preprocessing includes feature value extraction and feature spectrum drawing.

[0012] Furthermore, in step S3, analysis and diagnosis include: The sampled data after data preprocessing is used for fault analysis and diagnosis, including visible light image diagnosis, partial discharge fault diagnosis, infrared diagnosis and sound diagnosis; Among them, visible light image diagnosis includes: Perform personnel behavior recognition on the visible light images collected by the sensor, and use the YOLOv5s algorithm model to identify the wearing of helmets, work clothes, and smoking behavior in the visible light images; Collect image data including the above behaviors, annotate the data, frame the target area of ​​each person, and mark the corresponding category information; The model training is completed by adjusting the training parameters of the YOLOv5s model. The trained model is used for real-time reasoning to monitor whether the target appears in the visible light image. The judgment condition is that when the probability value of the helmet and work clothes is less than 75%, it means that the helmet is not worn and the work clothes are worn. When the probability value of smoking is greater than 75%, it means that smoking behavior exists, and an alarm is automatically generated; Partial discharge fault diagnosis, including: Use typical partial discharge feature maps as training data to build a classifier for the CNN convolutional neural network model; After the features of the spectrum data to be analyzed are extracted through the network, the average cosine similarity is calculated with the features of the spectrum library of various partial discharge types; After normalization, the similarity is used as the probability value of the four discharge types; If the probability value of a certain type exceeds 80% and the type is not external interference, an early warning is triggered, indicating that the device may have a partial discharge fault of this type; Infrared diagnostics, including: After infrared sampling, the temperature frame data is obtained, and the temperature dot matrix value is obtained after the relevant data conversion, filtering and compensation processing through multiple temperature frames; On the one hand, the partition monitoring temperature is extracted through the temperature dot matrix, and then compared with the preset threshold to analyze and judge the partition monitoring results; On the other hand, the thermal image is generated through the temperature dot matrix, and then processed by the image YOLOv5s algorithm to analyze and diagnose the results of the thermal image; Finally, the final conclusion of infrared diagnosis is obtained by combining the temperature analysis results of the partition monitoring and the diagnosis results of the thermal imaging diagram; Sound diagnostics, including: Collect the audio .wav files of different sound states of the cable under different environmental conditions collected by the cable sound sensor; Use MFCC Mel-frequency cepstral coefficient extraction method to extract the features of audio files and build an artificial neural network model for training; Deploy the trained model to the actual application scenario, receive audio input in real time, extract MFCC features and input them into the ANN model to diagnose and classify the sound into four levels; The model inference results are normalized, and the maximum confidence is used as the sound state category to realize the corresponding level warning function of the system.

[0013] Further, in step S4, the data storage includes: The data after analysis and diagnosis includes feature values, feature maps and diagnosis results. The structured data is stored locally through the database, and the feature maps are image files that are stored locally through file management.

[0014] Compared with the prior art, the multi-mode data fusion fault monitoring device and method described in the present invention have the following advantages: The present invention discloses a multi-mode data fusion fault monitoring device and method, which can integrate infrared, image, sound and partial discharge detection functions, and realize multi-dimensional monitoring of substation equipment. The present invention adopts machine learning algorithms and intelligent diagnosis technology to improve the accuracy and sensitivity of fault identification. The pan-tilt design of the present invention enables the device to have all-round monitoring capabilities and reduces monitoring blind spots. The distributed installation method of the present invention improves the scalability and deployment flexibility of the device. The present invention realizes real-time early warning and can effectively shorten the response time of equipment failure. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] The accompanying drawings constituting a part of the present invention are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the accompanying drawings: Figure 1 TOP schematic diagram of the substation online monitoring system according to an embodiment of the present invention; Figure 2 A schematic diagram of a monitoring device according to an embodiment of the present invention; Figure 3 A partial schematic diagram of a monitoring device according to an embodiment of the present invention; Figure 4 A schematic diagram of a pan / tilt platform according to an embodiment of the present invention; Figure 5 A schematic diagram of the hardware principle of the monitoring device according to an embodiment of the present invention; Figure 6This is a circuit diagram of a sound collection module according to an embodiment of the present invention; Figure 7 This is a schematic diagram of a LoRa communication module circuit according to an embodiment of the present invention; Figure 8 A schematic diagram of the software principle of the monitoring device according to an embodiment of the present invention; Fig. 9 A schematic diagram of the main business process of the fault monitoring method according to an embodiment of the present invention; Fig.10 This is a schematic diagram of the data acquisition module flow according to an embodiment of the present invention; Fig.11 This is a schematic diagram of the visible light collection principle flow chart according to an embodiment of the present invention; Fig.12 This is a schematic diagram of the principle flow of the data management module according to an embodiment of the present invention; Fig.13 This is a schematic diagram of the visible light image diagnosis process according to an embodiment of the present invention; Fig.14 A schematic diagram of a partial discharge fault diagnosis process according to an embodiment of the present invention; Fig.15 This is a schematic diagram of the infrared diagnosis process according to an embodiment of the present invention; Fig.16 The figure is a schematic diagram of the sound diagnosis process according to an embodiment of the present invention. DETAILED DESCRIPTION

[0016] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments may be combined with each other.

[0017] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", and the like are only used for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, features defined as "first", "second", and the like may explicitly or implicitly include one or more of the features. In the description of the present invention, unless otherwise specified, "multiple" means two or more.

[0018] In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood by specific circumstances.

[0019] The present invention will be described in detail below with reference to the accompanying drawings and in conjunction with embodiments.

[0020] like Figure 2 As shown, a multi-mode data fusion fault monitoring device, the monitoring device is mainly composed of two major components. One is the perception terminal, which mainly integrates various monitoring modules and is responsible for data collection, processing, analysis, storage and upload. The other component is the pan-tilt, which is responsible for the angle and direction rotation of the device. The two components are rigidly connected through the pan-tilt quick-release plate, and the perception terminal controls the rotation of the pan-tilt through the RS485 communication line. Specifically, the pan-tilt quick-release plate connects the perception terminal and the pan-tilt through M4 screws, and the perception terminal is installed with an infrared sensor. The sound sensor and light sensor are installed on both sides of the infrared sensor. The visible light camera and infrared fill light are also installed on both sides of the infrared sensor. Three indicator lights are installed between the visible light camera and the infrared fill light. A grounding screw hole is also provided on one side of the perception terminal. The bottom of the perception terminal is connected to the pan-tilt through the perception terminal network port connector and the pan-tilt network port cable. The bottom of the perception terminal is also connected to the pan-tilt through the pan-tilt connector male head and the pan-tilt PIN line. A pan-tilt network cable is also provided on one side of the pan-tilt. One end of the pan-tilt network cable is connected to the RJ45 connector extender, and one side of the pan-tilt is connected to the switch power supply through the pan-tilt power cable.

[0021] The monitoring device is mainly used for online monitoring of substations and is the core of the substation online monitoring system. Figure 1 It is the TOP diagram of the entire substation online monitoring system. The substation online monitoring system includes the main station online monitoring system, routers, several monitoring devices, power equipment GIS, power equipment main transformer, power equipment switchgear and power equipment high-voltage reactor. Several monitoring devices are connected to the main station online monitoring system through routers. Several monitoring devices are used to monitor the power equipment GIS, power equipment main transformer, power equipment switchgear and power equipment high-voltage reactor.

[0022] In this embodiment, the monitoring device can conduct comprehensive monitoring, analysis and diagnosis of the main power equipment of the substation (such as power equipment GIS, power equipment main transformer, power equipment switch cabinet and power equipment high-voltage reactor, etc.) by image (visible light), infrared, partial discharge and sound. And upload the sampling data and diagnosis results to the online monitoring system of the main station through wired or wireless means.

[0023] like Figure 3 As shown in the figure, the sensing terminal is the core component of the monitoring device, which integrates a visible light camera, an infrared thermoforming camera, a night infrared fill light, an ambient sound sensor, a light sensing sensor, a partial discharge sensor, and a temperature and humidity sensor. It is mainly responsible for realizing multi-dimensional information collection of power equipment, and processing, storing, analyzing and diagnosing the collected data, and uploading alarm information and sampling data in a timely manner.

[0024] like Figure 4 As shown in the figure, the gimbal is rigidly connected to the sensing terminal through the gimbal quick release plate. The gimbal is responsible for completing the angle and direction rotation of the device. The rotation of the gimbal is controlled by the sensing terminal. The horizontal rotation of the gimbal supports continuous rotation from 0 to 360 degrees, and the pitch angle is +60 to -60 degrees.

[0025] like Figure 5 As shown, the sensing terminal includes a main control module, a power module, a visible light image acquisition module, an infrared acquisition module, a partial discharge acquisition module, a sound acquisition module, a temperature and humidity acquisition module, a data transmission module, a security module, a data storage module and an input and output module.

[0026] In this embodiment, the sensing terminal mainly includes: Main control module: The main control module of the sensing terminal adopts ARM A7 quad-core processor to realize the scheduling of various modules.

[0027] Power module: including high voltage terminal, 220VAC to 5V, DCDC module, POE module, etc. Realize the power supply function for the sensing terminal components.

[0028] Visible light image acquisition module: mainly includes high-resolution visible light camera, infrared fill light, light sensor (light sensing sensor) and IRCUT (infrared filter). It is used to monitor equipment appearance defects (such as cracks, oil stains, etc.) and image recognition detection (such as personnel intrusion, dress code recognition, etc.). The light sensor (light sensing sensor) monitors the light brightness around the equipment, and links IRCUT and infrared fill light to assist the visible light camera to complete image acquisition.

[0029] Infrared acquisition module: mainly includes infrared sensors (infrared cameras). It is used to collect temperature information of substation equipment and generate thermal images to detect temperature rise and infrared anomalies of power equipment.

[0030] Partial discharge acquisition module: mainly includes partial discharge antenna, partial discharge sensor and conditioning circuit. It is used to monitor the partial discharge signal of power equipment and judge the internal insulation status of the equipment.

[0031] Sound collection module: mainly includes sound sensors. It is used to monitor the movement of mechanical parts of power equipment or the soundprint characteristics of the body to determine whether it is abnormal. The circuit diagram of the sound collection module is shown in Figure 6 .

[0032] Temperature and humidity acquisition module: mainly includes temperature and humidity sensors. It is used to monitor the ambient temperature and humidity around the equipment, and links with intelligent diagnosis algorithms to achieve comprehensive analysis and diagnosis.

[0033] Data transmission module: including wired communication module and wireless communication module, mainly responsible for uploading and receiving data and instructions. The wired communication module is mainly composed of RJ45, network card chip, etc.; the wireless communication module mainly includes 4G communication module (mainly composed of SIM card, 4G antenna), LoRa communication module (mainly composed of LoRa module, LoRa antenna), and WiFi / BT / GPS three-in-one module (mainly composed of WiFi / BT / GPS, WiFi / BT / GPS antenna). The LoRa communication circuit diagram is shown in Figure 7 .

[0034] Security module: mainly includes encryption chip, responsible for data encryption and decryption.

[0035] Data storage module: mainly includes TF card, responsible for the storage and management of local data, pictures and files.

[0036] Input and output module: mainly includes indicator lights to display the operating status, power supply status and communication status of the sensing terminal.

[0037] In this embodiment, the pan / tilt includes: PTZ module: and PTZ body, used to support the entire device and achieve multi-angle and all-round equipment monitoring.

[0038] Switching power supply: provides power to the gimbal.

[0039] like Figure 8 As shown, this is a software system framework diagram of the monitoring device. The software system adopts a modular design and is deployed in the sensing terminal component of the monitoring device.

[0040] The software system mainly consists of: data acquisition module, data management module, security module, data transmission module, input and output module, power supply module and aggregation module.

[0041] Data acquisition module (including visible light image acquisition module, infrared acquisition module, partial discharge acquisition module, sound acquisition module, temperature and humidity acquisition module): mainly responsible for the acquisition of sensor data such as visible light, partial discharge, infrared, sound, temperature and humidity. It is also responsible for dispatching GPS and PTZ to complete positioning and rotation of monitoring angles.

[0042] Data management module (data storage module): mainly responsible for the preprocessing, analysis and diagnosis of sampled data, as well as the storage of related data and files. After the data management module makes real-time analysis and diagnosis of the status of the monitored equipment, it issues a warning in time (when the diagnosis result is abnormal) through the data upload subsystem (external communication interface 4G or RJ45), and regularly uploads the sampled data of each sensor.

[0043] Security module: mainly responsible for data encryption and decryption to ensure secure data transmission.

[0044] Data transmission module: mainly responsible for data encapsulation and parsing, and completes the sending and receiving of data and instructions through wired and wireless methods.

[0045] Input / output module: mainly used for user interaction of the device, such as indicator lights (system status, power status, communication status indication), buttons (power on / off), etc.

[0046] Power module: mainly used to ensure the power supply and power consumption control of the system.

[0047] Aggregation module: the aggregation node function of the sensing terminal, which expands and realizes the data access function of other sensor devices through LoRa.

[0048] like Fig. 9 As shown in the figure, a multi-mode data fusion fault monitoring method has the following main business processes: The main function of the monitoring device is to collect and analyze the data of various dimensions of power equipment, monitor its working status and provide timely warning. The process mainly includes data acquisition module, data management module, security module and data transmission module.

[0049] like Fig.10 As shown, the data acquisition module mainly includes sampling scheduling, pan-tilt control, GPS information acquisition and sensor multi-channel sampling functions.

[0050] Sampling scheduling is the core function of the sampling room module, which initiates sampling instructions at regular intervals and controls the pan-tilt rotation, GPS acquisition, and sensor multi-channel acquisition in turn.

[0051] (1) PTZ control means controlling the PTZ components to rotate according to the preset positions (horizontal angle and pitch angle); (2) GPS collection is to collect the latitude and longitude and location information of the monitoring device; (3) After the PTZ rotates to the specified preset position, the multi-channel sampling of the sensor is started, including data collection of visible light, partial discharge, infrared, sound, temperature and humidity.

[0052] like Fig.11 As shown, visible light collection: Visible light collection mainly involves light sensors, visible light cameras, infrared fill lights and IRCUT.

[0053] First, the light sensor (light sensing sensor) analyzes the brightness of the light around the device, and then links the IRCUT (infrared filter) and infrared fill light to assist in the collection of visible light. When the light sensor detects a dark scene, it will first turn on the IRCUT, then turn on the infrared fill light, and finally call the visible light camera to take pictures, ensuring the photo effect in dark light and facilitating subsequent image analysis. The detailed process is shown in the figure below: like Fig.12 As shown, the data management module includes three parts: data preprocessing, analysis and diagnosis, and data storage.

[0054] Data preprocessing includes: The visible light, partial discharge, infrared, and sound sensor data collected by the data acquisition module need to be pre-processed before analysis and diagnosis can be performed.

[0055] Visible light image preprocessing includes distortion processing, brightness and darkness optimization, etc., to generate high-quality images for subsequent analysis.

[0056] The preprocessing of partial discharge data includes eigenvalue extraction and feature spectrum drawing to facilitate subsequent analysis.

[0057] Infrared data preprocessing includes filtering and compensation processing of temperature frame data (temperature dot matrix), extraction of temperature extreme value for partition monitoring, and generation of infrared thermal imaging images.

[0058] The preprocessing of sound data includes feature value extraction, feature spectrum drawing, etc.

[0059] Analytical diagnosis includes: The pre-processed sampled data (including eigenvalues, characteristic spectra, etc.) can be used for fault analysis and diagnosis. Edge algorithms such as machine learning are used to perform abnormal diagnosis on visible light images, partial discharge data, infrared data, and sound data to identify potential faults and fault types.

[0060] (1) Visible light imaging diagnosis: The images collected by the sensor are used for personnel behavior recognition, and the YOLOv5s algorithm model is used to identify the wearing of helmets, work clothes, smoking, and other behaviors in visible light images. Image data containing these behaviors is collected and annotated. The target area of ​​each person is framed and the corresponding category information (helmet, work clothes, smoking) is marked. The model training is completed by adjusting the YOLOv5s model training parameters. The trained model is used for real-time reasoning to monitor whether the target (helmet, work clothes, smoking) appears in the visible light image. The judgment condition is that when the probability value of helmets and work clothes is less than 75%, it means that the helmet is not worn and work clothes are worn. When the probability value of smoking is greater than 75%, it means that smoking behavior exists. Alarms are automatically generated to ensure safety supervision, such as Fig.13 shown.

[0061] (2) Partial discharge fault diagnosis: Using typical partial discharge feature maps as training data, a classifier of the CNN convolutional neural network model is constructed. The convolution layer and fully connected layer of the classifier can effectively extract the features of the map data. After the map data to be analyzed is extracted through the network, the average cosine similarity is calculated with the map library features of various partial discharge types. After the similarity is normalized, it is used as the probability value of four discharge types (external interference, tip discharge, insulation discharge and suspended discharge). If the probability value of a certain type exceeds 80% and the type is not external interference, an early warning is triggered, indicating that the equipment may have a fault of this type of partial discharge, such as Fig.14 shown.

[0062] (3) Infrared diagnosis: After infrared sampling, the temperature frame data (temperature dot matrix) is obtained. Through the conversion of relevant data in multiple temperature frames, filtering and compensation processing, the accurate temperature dot matrix value is obtained. On the one hand, the partition monitoring temperature is extracted through the temperature dot matrix, and then compared with the preset threshold to analyze and judge the partition monitoring results; on the other hand, the thermal imaging image is generated through the temperature dot matrix, and then processed by the image YOLOv5s algorithm to analyze and diagnose the results of the thermal imaging image. Finally, the final conclusion of infrared diagnosis is obtained by combining the partition monitoring temperature analysis results and the thermal imaging image diagnosis results, such as Fig.15 shown.

[0063] (4) Sound diagnosis: The audio .wav files of different sound states of the cable under different environmental conditions collected by the cable sound sensor are collected. The features of the audio files are extracted using the MFCC Mel-frequency cepstral coefficient extraction method. An artificial neural network model is constructed for training. The trained model is deployed to the actual application scenario. The audio input is received in real time. After MFCC feature extraction, it is input into the ANN model for diagnosis and classification of the four sound status levels (normal state, attention state, serious state, and critical state). The model inference results are normalized, and the maximum confidence is used as the sound status category to realize the corresponding level warning function of the system, such as Fig.16 shown.

[0064] Data storage includes: The data after analysis and diagnosis includes feature values, feature maps, and diagnosis results. The structured data is stored locally through the database, and the feature maps are image files that are stored locally through file management.

[0065] The monitoring device can be installed in a distributed manner near the substation equipment. Each device can operate independently, or multiple devices can work together through the network.

[0066] Advantages of the present invention: The present invention can integrate infrared, image, sound and partial discharge detection functions to achieve multi-dimensional monitoring of substation equipment. The present invention adopts machine learning algorithms and intelligent diagnosis technology to improve the accuracy and sensitivity of fault identification. The pan-tilt design of the present invention enables the device to have all-round monitoring capabilities and reduces monitoring blind spots. The distributed installation method of the present invention improves the scalability and deployment flexibility of the device. The present invention achieves real-time early warning and can effectively shorten the response time of equipment failures.

[0067] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A multi-mode data fusion fault monitoring device, characterized in that: It includes a sensing terminal and a pan-tilt head. The pan-tilt head is installed at the bottom of the sensing terminal through a pan-tilt head quick-release plate. A switching power supply is installed on one side of the pan-tilt head. The sensing terminal is connected to the pan-tilt head through an RS485 communication line. The hardware structure of the perception terminal includes a main control module, a power module, a visible light image acquisition module, an infrared acquisition module, a partial discharge acquisition module, a sound acquisition module, a temperature and humidity acquisition module, a data transmission module, a security module, a data storage module and an input and output module, and the power module, the visible light image acquisition module, the infrared acquisition module, the partial discharge acquisition module, the sound acquisition module, the temperature and humidity acquisition module, the data transmission module, the security module, the data storage module and the input and output module are all communicatively connected with the main control module; The pan-tilt head comprises a switching power supply and a pan-tilt head module. The switching power supply is used to supply power to the pan-tilt head module, and the pan-tilt head module is communicatively connected with the main control module.

2. A multi-mode data fusion fault monitoring device according to claim 1, characterized in that: The main control module adopts ARM A7 quad-core processor to realize the scheduling of each module; The power module includes a high-voltage terminal, a 220VAC to 5V module, a DCDC module and a POE module. The high-voltage terminal is connected to the DCDC module through the 220VAC to 5V module, the POE module is also connected to the DCDC module, and the DCDC module is connected to the main control module; The visible light image acquisition module includes a visible light camera, an infrared fill light, a light sensing sensor and an infrared filter, all of which are connected to the main control module for communication; The infrared acquisition module includes an infrared sensor, which is used to collect temperature information of the substation equipment and generate a thermal image; The partial discharge acquisition module includes a partial discharge antenna, a partial discharge sensor and a conditioning circuit. The input end of the conditioning circuit is connected to the main control module for communication. The conditioning circuit is connected to the partial discharge antenna through the partial discharge sensor. The sound collection module includes a sound sensor that is communicatively connected to the main control module; The temperature and humidity acquisition module includes a temperature and humidity sensor that is communicatively connected to the main control module; The data transmission module includes a wired communication module and a wireless communication module. The wired communication module is mainly composed of RJ45, a network card chip, etc. The wireless communication module includes a 4G communication module, a LoRa communication module, and a WiFi / BT / GPS three-in-one module. The security module includes an encryption chip that is communicatively connected to the main control module; The data storage module includes a TF card that is communicatively connected to the main control module; The input-output module includes an indicator light that is communicatively connected to the main control module.

3. The multi-mode data fusion fault monitoring device according to claim 1, characterized in that: The software structure of the perception terminal includes a data acquisition module, a data management module, a security module, a data transmission module, an input / output module, a power module and a convergence module. The data management module is communicated with the data acquisition module and the security module respectively. The security module is also communicated with the data transmission module and the main control module respectively. The input / output module, the power module and the convergence module are all communicated with the main control module.

4. The multi-mode data fusion fault monitoring device according to claim 3 is characterized in that: The visible light image acquisition module, infrared acquisition module, partial discharge acquisition module, sound acquisition module, and temperature and humidity acquisition module constitute a data acquisition module, which is responsible for collecting sensor data, dispatching GPS and the pan-tilt head, completing positioning, and rotating the monitoring angle; The data management module is responsible for the preprocessing, analysis and diagnosis of sampled data, as well as the storage of related data and files; The security module is responsible for data encryption and decryption to ensure secure data transmission; The data transmission module is responsible for data encapsulation and analysis, and completes the sending and receiving of data and instructions through wired and wireless means; The input-output module is used for user interaction of the device; The power supply module is used to ensure the power supply and power consumption control of the system; The aggregation module expands and implements the data access function of other sensor devices through LoRa.

5. A multi-mode data fusion fault monitoring method, using the multi-mode data fusion fault monitoring device according to any one of claims 1 to 4, characterized in that: The following steps are involved: S1, data collection based on data collection module; S2, performing data preprocessing on the data processed in step S1 based on the data management module; S3, analyzing and diagnosing the data processed in step S2 based on the data management module; S4, storing the data processed in step S3 based on the data management module; S5. Encrypt and decrypt the data processed in step S4 based on the security module; S6. Push an alarm for the data processed in step S5 based on the data transmission module.

6. A multi-mode data fusion fault monitoring method according to claim 5, characterized in that: In step S1, data collection includes: S11, sampling scheduling, and judging whether sampling is required, if yes, proceed to step S12 and step S13 at the same time, if no, continue sampling scheduling; S12, controlling the pan / tilt component to rotate according to the preset position, and proceeding to step S14; S13, collecting the latitude and longitude and location information of the monitoring device; S14, start multi-channel sampling of sensors; The sensor's multi-channel sampling includes visible light collection, partial discharge collection, infrared collection, sound collection, and temperature and humidity collection.

7. A multi-mode data fusion fault monitoring method according to claim 6, characterized in that: In step S14, visible light collection includes: First, the light sensor is used to analyze the brightness of the light around the device, and the infrared filter and infrared fill light are linked to assist in the collection of visible light; When the light sensor detects a dark scene, first turn on IRCUT, then turn on the infrared fill light, and finally call the visible light camera to take pictures.

8. The multi-mode data fusion fault monitoring method according to claim 5, characterized in that: In step S2, data preprocessing includes visible light image preprocessing, partial discharge data preprocessing, infrared data preprocessing and sound data preprocessing; Among them, visible light image preprocessing includes distortion processing and brightness optimization; Partial discharge data preprocessing includes eigenvalue extraction and feature map drawing; Infrared data preprocessing includes filtering and compensation processing of temperature frame data, extraction of temperature extreme value for zone monitoring, and generation of infrared thermal imaging images; Sound data preprocessing includes feature value extraction and feature spectrum drawing.

9. The multi-mode data fusion fault monitoring method according to claim 5, characterized in that: In step S3, analysis and diagnosis include: The sampled data after data preprocessing is used for fault analysis and diagnosis, including visible light image diagnosis, partial discharge fault diagnosis, infrared diagnosis and sound diagnosis; Among them, visible light image diagnosis includes: Perform personnel behavior recognition on the visible light images collected by the sensor, and use the YOLOv5s algorithm model to identify the wearing of helmets, work clothes, and smoking behavior in the visible light images; Collect image data including the above behaviors, annotate the data, frame the target area of ​​each person, and mark the corresponding category information; The model training is completed by adjusting the training parameters of the YOLOv5s model. The trained model is used for real-time reasoning to monitor whether the target appears in the visible light image. The judgment condition is that when the probability value of the helmet and work clothes is less than 75%, it means that the helmet is not worn and the work clothes are worn. When the probability value of smoking is greater than 75%, it means that smoking behavior exists, and an alarm is automatically generated; Partial discharge fault diagnosis, including: Use typical partial discharge feature maps as training data to build a classifier for the CNN convolutional neural network model; After the features of the spectrum data to be analyzed are extracted through the network, the average cosine similarity is calculated with the features of the spectrum library of various partial discharge types; After normalization, the similarity is used as the probability value of the four discharge types; If the probability value of a certain type exceeds 80% and the type is not external interference, an early warning is triggered, indicating that the device may have a partial discharge fault of this type; Infrared diagnostics, including: After infrared sampling, the temperature frame data is obtained, and the temperature dot matrix value is obtained after the relevant data conversion, filtering and compensation processing through multiple temperature frames; On the one hand, the partition monitoring temperature is extracted through the temperature dot matrix, and then compared with the preset threshold to analyze and judge the partition monitoring results; On the other hand, the thermal image is generated through the temperature dot matrix, and then processed by the image YOLOv5s algorithm to analyze and diagnose the results of the thermal image; Finally, the final conclusion of infrared diagnosis is obtained by combining the temperature analysis results of the partition monitoring and the diagnosis results of the thermal imaging diagram; Sound diagnostics, including: Collect the audio .wav files of different sound states of the cable under different environmental conditions collected by the cable sound sensor; Use MFCC Mel-frequency cepstral coefficient extraction method to extract the features of audio files and build an artificial neural network model for training; Deploy the trained model to the actual application scenario, receive audio input in real time, extract MFCC features and input them into the ANN model to diagnose and classify the sound into four levels; The model inference results are normalized, and the maximum confidence is used as the sound state category to realize the corresponding level warning function of the system.

10. The multi-mode data fusion fault monitoring method according to claim 5, characterized in that: In step S4, data storage includes: The data after analysis and diagnosis includes feature values, feature maps and diagnosis results. The structured data is stored locally through the database, and the feature maps are image files that are stored locally through file management.

Images