Deep Learning Based Power Switch Detection And State Recognition

As the size of power networks continues to expand,the safe operation of power systems becomes more and more important.As an important part of the power grid,substations are one of the key factors that determine whether the grid can operate safely.With the popularization and application of video surveillance technology,the unattended substation has gradually become the possible future of power grid dispatch automation.In order to guarantee the normal operation of the substation,it is necessary to monitor the on and off state of the switches in substations.Therefore,automatic detection and status recognition of switch is the technical basis for unattended substations and also an important guarantee for the safe operation of substations.This Research report introduces a research work on automatic detection and state recognition of 220 KV switch based on computer vision technology.We first proposed a deep learning based switch detection algorithm,where the Faster R-CNN is used to detect the switching slots formed by two rings;then ROI of the switch rod position is automatically extracted based on prior knowledge;finally,a SVM classifier is trained to identify whether there is a switch rod in the ROI to determine the status of the switch.The prototype system we developed has good performance.The processing speed is only 4 to 5 seconds per image on a CPU,and the detection accuracy is 99%.