Abstract
With the evolution of the electric power industry, maintenance and monitoring technologies for substations are continuously being innovated. Among these, oil sampling robots hold significant potential for application and are expected to emerge as a research hotspot in the field of intelligent robotics. However, the existing technology for oil extraction robots is not sufficiently refined, and their application outcomes are not as ideal as desired. This paper proposes a fully autonomous oil sampling robot system for routine maintenance operations of oil-filled equipment within substations, aiming to enhance operational safety, sampling efficiency, and sample accuracy, while reducing manual operational costs. The system achieves full terrain adaptability in complex environments, oil path sealing, stable operation capabilities, and fully automatic docking with oil outlets through a tracked mobile chassis, sealed oil passageways, a gantry-style robotic arm, and Aruco marker positioning technology. Furthermore, this paper introduces a path planning algorithm based on Deep Attention Q-Network to address the issue of robot path planning in partially observable environments. The effectiveness of the proposed method is verified through comparative experiments. The robot has been successfully applied in a demonstration of oil sampling for oil-filled equipment at a 1000 kV ultra-high voltage substation, thereby validating the reliability of the robot's structure and functionality.