Substation intelligent operation and maintenance and scheduling system
In the regulation and integration mode, the control personnel need to identify and confirm the execution status of the switches, knife gates, meters, etc. before and after the equipment operation. At present, a large number of equipment status, meter and site situation confirmation work still need to be completed by traditional CCTV manual identification and manual to the scene. For the method of remote video surveillance using environmental monitoring cameras, the actual use is affected by factors such as illumination, installation location, camera field of view, and imaging quality. It is difficult to effectively realize remote video confirmation, and effective automatic identification cannot be realized. However, due to the large number of traditional video surveillance points, high equipment costs, and large maintenance workload, many cameras are difficult to use in actual use.
In the case of abnormal equipment and bad weather, it is important to obtain the operating status of the equipment in a timely and accurate manner. If personnel cannot arrive at the site to collect on-site information and confirm the status of the equipment in time, it will inevitably lead to the extension of the accident disposal time, and may also lead to the expansion of the accident and affect the safe and stable operation of the power grid.
Therefore, it is urgent to develop a system that can realize the intelligent monitoring of the remote control of the substation in the integrated control mode, so as to realize the automatic identification of the switch and the position of the knife gate, the automatic reading of the instrument value/indication, and the intelligent state of the field analysis.
1.2 Product Profile
Through the establishment of a remote vision intelligent identification system, GMI achieves the requirement of integration of regulation and control, intelligently identifies and confirms the status of devices with remote operation conditions, such as switches, knives and gates, and reduces the outage time of equipment; to a certain extent, it improves the accident handling ability of dispatchers, reduces the economic losses caused by the upgrade of events, the expansion of faults and the failure of power grid; Control integration process, improve the intelligence of power grid operation, to a certain extent, reduce the workload of patrol personnel.
This system uses intelligent patrol platform, loads full spectrum recognition instrument and high definition network camera, collects scene data, carries out data analysis and summary through back-end centralized control platform, provides intelligent and automated substation patrol tasks, and reports and stores all data indicators of substation after analysis and summary.
1.3 Product Networking Architecture
This scheme adopts the hardware system based on machine vision technology, integrates the innovative software platform, combines the large data, Internet of Things, artificial intelligence and other technologies, deploys the intelligent multi-spectral acquisition platform, forms the information network covered by the whole network, and establishes an "integrated, visualized and intelligent" intelligent vision system.
The system adopts a hierarchical structure of front-end and back-end, in which the front-end is divided into linkage end and client end. The linkage terminal is responsible for isolating the control system, recognizing and responding to the control signal while avoiding affecting the existing system, while the client is responsible for presenting the system's operation status and results; the back end is divided into service terminal and acquisition terminal mobile platform running in the pilot station, which can realize the status acquisition, identification and statistics of field equipment through multi-spectral and visual algorithms.
The main facilities of the system are:
● Multispectral Mobile Information Acquisition Platform
1. Acquisition:based on day blind ultraviolet, thermal imaging, visible light to achieve knife gate closure identification;
2. Reading meters:2.Intelligent recognition and statistics of equipment status and related meters are realized by deep customized intelligent vision algorithm.
3. Processing:3.Superposition of ultraviolet and visible images, increase information labeling and improve identification;
4. Motion:4.Laser and vision technology are integrated through VSLAM navigation technology to realize autonomous positioning and navigation of intelligent vision system, which is the key to autonomous operation of mobile platform.
● Wireless Network Graphic Transmission System
It consists of a wireless AP and a wireless bridge, and forms a global wireless network within the site to ensure that data is not lost and interacts in real time. At the same time, the system comes with information encryption measures to fully realize the use of key information security protection for power grid pilots.
● Autonomous return charging system
Considering the unmanned maintenance of unattended substation smart equipment, we add an automatic return charging system for the mobile platform. The charging room is a movable board room, and after being easily installed, it can be moved to a suitable position in the pilot environment, eliminating the need for repetition. Civil construction, specializing in the construction of a charging room, rail and other facilities.
1.4 Programme features
The system is based on innovative software design, combined with power system dedicated network, vision networking technology, positioning technology, video compression technology, computer processing technology, etc., is a complete, efficient, cost-effective regulation-based integration Intelligent vision system.
● Multispectral visual inspection and recognition
Visible light has high resolution, intuitive image, easy to obtain texture, structure, motion, color, and spatial information, but ordinary pattern recognition and behavior analysis methods are difficult to find fault locations in time. Infrared/thermal imaging easily finds faults that have failed and generates hot and high temperatures based on temperature gradients and temperature field distributions, but not only thermal imaging has low resolution, high cost, but also it is difficult to find faults early in the problem. High-voltage discharge, arc, etc. are prone to invisible ultraviolet light. Ultraviolet light can more quickly locate fault points. However, special optical filters and algorithms are needed to eliminate natural light interference. In addition, ultraviolet light is weak and difficult to detect. Level amplification and complex noise reduction processing, and UV imaging positioning is very difficult. Combined with the characteristics of multi-spectral, it is possible to achieve rapid positioning of fault points, and combined with 3D control, it is easy to achieve accurate remote visualization.
● Fusion of UV imaging and visible light images and real-life enhancement
The function of the UV imager is to detect corona discharge and surface partial discharge that are invisible to the human eye outside the visible range. The spatial distribution and luminescence intensity of partial discharge (corona discharge) reflect the electric field distribution of the charged equipment, and the shortage of electric field anomalies can be found. However, due to the low resolution of the ultraviolet image, it is difficult to judge the image content from the ultraviolet image. Therefore, it is necessary to use the attitude information of the ultraviolet camera and the visible light camera to superimpose the ultraviolet image onto the visible light image through the affine transformation projection, and the user can directly The discharge phenomenon of the device is observed on the high-definition image, and the searching and positioning of the discharge device is facilitated.
● Multi-sensor data fusion machine vision system
Multi-sensor information fusion refers to the simultaneous use of multiple sensors of the same type or different types, synthesizing the partial incomplete observations and related data provided by each sensor, eliminating the possible redundancy and contradiction between the sensor information, and Complement each other to reduce its uncertainty and obtain a consistent description of the object or environment. The accuracy of data fusion determines the pose estimation, the matching accuracy of the map feature points, and can usually be extracted by laser radar or visual processing systems (such as Kinect, monocular camera, binocular camera, etc.), and then matched from the feature level. Fusion.
● Intelligent front machine vision system
The increase in the resolution of the visual sensor leads to a huge amount of data, which poses a challenge to both data transmission and processing. Data compression is an effective video/visual data transmission processing method, but compression distortion will cause problems in pattern recognition and 3D positioning of the back end. The distributed intelligent front vision system not only can efficiently process the intelligent analysis/pattern recognition tasks of each front end, but also has high stability and no dependence on the back-end servers. There is no direct resource conflict between the front ends, and the system is more stable.
● Isolated production scheduling system captures the linkage mechanism of remote knife gate identification system
In order to realize the linkage between the production scheduling system and the system, and at the same time to ensure the safety of the production scheduling system, the system realizes an isolated linkage mechanism, that is, the display of the production scheduling system is captured by the visual sensor in the case of physical isolation of the network. Through the machine vision and pattern recognition algorithm to read the image semantics, to send instructions to the identification system, control the corresponding surveillance camera and mobile platform to identify operations, read the relevant status of the knife gate, switch and other devices for cross-validation, and scheduling instructions form a closed loop.