Automated control of switchings operations in electrical complexes and systems based on machine vision

Ensuring safety during operational switching in electrical installations is a critically important task. The current regulatory requirements provide for the use of mechanical and electrical interlocks, but there remains a risk of circumvention or malfunction, which can lead to serious consequences. The article discusses an approach to creating an additional, independent control tool based on machine vision technology. The object of the study was the process of operational switching and repair work on the cells of complete switchgear (CRU) with a voltage of 10 kV. The purpose of the work is to develop and experimentally test a prototype system for automated control of the correctness of personnel actions. The system includes a camera mounted on a worker's helmet, a control panel image stabilization algorithm based on fiducial markers, a neural network element detection module (based on the YOLOv11 architecture) and a logical module for analyzing the sequence of actions in real time. The paper considers the sequence of operations that are necessary to repair the control panel switch, as well as to return it to its working position after repair. In the course of experimental research, the YOLOv11m and YOLOv11l models were trained on their own labeled dataset of control panel images. The best recognition accuracy (up to 97%) was achieved for the condition of the light indicators, satisfactory accuracy (78-84%) - for detecting the fact of pressing the button. The testing of a software and hardware prototype integrated with a modified helmet with an electronic unit based on Arduino Nano has confirmed the operability of the approach. The article presents an analysis of the obtained results, as well as a plan for further work aimed at improving the accuracy and speed of the system for its subsequent implementation as a means of duplicate security control.