Deep learning-based safety behavior identification of operations and maintenance personnel on a substation

Abstract

With the growth of the economy, electricity consumption is also increasing. To ensure the regular running of substations, operation and maintenance personnel need to inspect the equipment on-site, but they are also exposed to a hazardous environment. Since the source of danger cannot be eliminated entirely, the safety behavior of the operation and maintenance personnel in the substation needs to be identified. For this purpose, we propose an automated approach based on deep learning. We first collected images of operation and maintenance personnel on a substation as a dataset. Then, an object detection model based on YOLO v5 is proposed, which can accurately detect the substation's operation and maintenance personnel, and the personal protective equipment (PPE). Combining with individual posture estimation, using a one-dimensional convolutional neural network (1D-CNN), it is able to determine whether the PPEs are being worn correctly. Finally, the safety behavior identification is verified by an experiment, and satisfactory results are obtained. With the method in this paper, it is possible to automatically identify whether substation operations and maintenance personnel are wearing PPE correctly according to the regulations, which has a positive impact on improving the efficiency of safety management of substations.