 This paper proposes an improved version of the GFYOLOV7 network for detecting the status of bounce locks in a substation. It uses the mobile Viteem module to enhance the feature extraction capabilities of the backbone network, while also incorporating the CBAM feature attention mechanism and two types of attention modules to further reduce the network size and increase its accuracy. The experimental results demonstrate that the proposed network achieves a detection accuracy of 97.8% with a reduced network weight and a slight decrease in detection speed. This makes it suitable for use in real-time applications such as substation monitoring. This article was authored by Yang Wang, Xiao Fengzhang, Long Meili, and others.