Abstract: The accurate identification and verification of the state of power equipment used in substation operations would allow intelligent substations to be operated unattended under severe weather and complex background conditions. Currently, the main functions of robot or online monitoring systems operated by substations are to provide human-assisted inspections, not automatic identification or calibration. Combined with the actual conditions of smart grid substations and construction requirements, this paper proposes an algorithm based on computer intelligent vision technology that can be used for the automatic identification of typical outdoor circuit breakers and disconnectors and the position state of indoor switchgears. Firstly, a scale invariant feature transform (SIFT) algorithm was used to accurately locate the area to be detected. Then, image preprocessing technology was used to remove noise points and extract edge information. The randomized Hough transform was used to extract the line information of the disconnector and the circle information of the switchgear, and k-NN (k nearest neighbor) was used to extract and identify the written character information on the circuit breaker. Finally, intelligent identification was set up using thresholds for three types of power equipment, and the algorithm was verified for a disconnector in a 500-kV substation in China and at the Qinghe substation. Based on actual measurements at these sites, the algorithm exhibited strong identification performance, stability, and high accuracy in identifying the position state of the switches in severe weather. It can be used to not only solve the current lack of effective automatic online monitoring of power equipment, but also serve as an important extension of many areas of smart grid research.

Key words: computer vision, intelligent robot, sift algorithm, randomized Hough transform, intelligent identification