Research and application of composite insulator overheat identification method based on improved case segmentation

• Published in: Journal of physics. Conference series 2024-08, Vol.2814 (1), p.12045
• Main Authors: Rao, Chengcheng, Liu, Gao, Li, Changyu
• Format: Article
• Language: English
• Subjects: Algorithms; Error detection; Identification methods; Image segmentation; Infrared heating; Infrared imagery; Infrared inspection; Instance segmentation; Insulators; Overheating; Position measurement; Power lines; Safety measures; Time measurement
• ISSN: 1742-6588; 1742-6596
• DOI: 10.1088/1742-6596/2814/1/012045

How to quickly and accurately locate the heat of power components is one of the problems that need to be discussed frequently in the power field. At the same time, due to the diverse types of equipment in the power industry and the complex environment, if the heating position of the components cannot be located in time and further safety measures are taken, it will greatly affect whether the line can run stably. Based on infrared images, the device can be detected at a distance, and the problem can be found without interrupting its working state, which is more efficient in the process of equipment inspection. At present, some algorithms are based on infrared image research, but the positioning is not accurate, and some problems such as the temperature extracted by the algorithm and the temperature deviation between the actual parts caused by environmental factors, have not been a perfect solution. With regard to these problems, this paper presents a new method for overheating identification of composite insulators. First, the external profile of composite insulators is identified in infrared images through the YOLOV5 instance segmentation algorithm, and the rod core position of composite insulators is located based on this external profile. Then, the temperature matrix of the rod core position is obtained through the original heat matrix in infrared images. By comparing the temperature of the temperature matrix, we can determine whether the temperature is abnormal. In the actual production environment, the accuracy rate of the proposed method is 0.84, the error detection ratio is 0.2, and the total time of the algorithm is 216 ms. The results show that the proposed method has a good component positioning effect and can filter most of the temperature identification errors caused by environmental factors. It provides a new solution to solve the problem of temperature anomaly identification of overhead line power equipment in different scenarios.