Identification of the closing position of scissor-type disconnectors based on a U-net network

The safe operation of power systems is an important support and guarantee for the development of national economy. Disconnectors are one kind of the most important equipments in an electricity substation. The detection of the closed state of a disconnector has become an important research topic in the field of power. This paper studies scissor-type disconnectors, which are used in new 220kV or 500kV electricity substations in recent years. The research direction of this paper is not to judge whether a disconnector is open or closed, but to focus on whether the disconnector is closed in place. In this paper, a U-net network, improved from FCN networks, which are suitable for small data sets and have high accuracy, is used to complete the task of semantic segmentation. The U-net network can separate the target from the surrounding environment, extract the outline of a disconnector, and avoid the interference of surrounding objects to the target.