针对电力系统作业人员作业过程中安全帽佩戴、抽烟及违规进入危险区域等行为所带来的安全风险问题，文中研究了基于改进YOLO( you only look once) v7的作业安全风险智能预警方法。论述了基于深度学习的新型电力系统作业风险预警系统模型训练和使用架构，分析了基于YOLO v7的目标识别和检测方法，提出了通过在特征提取中引入卷积注意力机制和通过k-means + +算法优化预选框的模型改进方法。最后利用改进的YOLO v7模型对前述作业风险预警效果进行了分析和验证。结果表明，相比于改进前的YOLO系列检测方法，所述改进模型在电力系统作业安全风险检测中具有更好的性能,能够满足新型电力系统对相关作业风险的实时检测和预警需求。
英文摘要:
Regarding the safety risks caused by improper wearing of safety helmets, smoking, and illegal entry into hazardous areas during the operation of power system, this paper studies an intelligent warning method for safety risks in operations based on improved YOLO (you only look once) v7 deep learning model. Firstly, the architecture for training and using a deep learning-based operation risk warning model is discussed. Then, the object recognition and detection method based on the YOLO v7 deep learning model is analyzed, and convolutional block attention module (CBAM) and k-means ++ methods are introduced to improve the model. Finally, the improved YOLO v7 model is used to analyze and verify its effectiveness in the aforementioned safety risk warning. Experimental results show that compared to the previous YOLO series detection methods, the improved YOLO v7 model has better performance in detecting operational safety risks in power system, and can meet the real-time detection and risk warning needs of the novel power system for operational risks.