Study of spectral image extraction and component identification of insulator surface contamination

Le résumé fourni par la source

The accumulation of dirt components on the surface of composite insulators is a significant factor influencing the dirt flashover voltage. In order to identify typical fouling components in a more efficient and practical manner, this paper proposes a fouling component identification method based on reflectance hyperspectral imaging. Firstly, a single or a mixed sample of seven typical dirt-accumulating silicone rubber samples were prepared by the standard artificial dyeing method, and the hyperspectral two-dimensional image element data of the dirt-accumulating samples were obtained under simulated daylight conditions. This was followed by the establishment of a three-dimensional feature subspace based on envelope elimination in the spectral dimension. Finally, the kernel support vector machine algorithm (KSVM) was introduced to achieve the image element identification of the seven fouling components. The results demonstrate that the accuracy of this method reaches 93.75% for the identification of the dirt components accumulated on the surface of silicone rubber, and that the accuracy is higher than 80% for the identification of the mixed dirt components. This study presents a novel approach for the rapid analysis of composite insulator filth composition and the evaluation of flashover characteristics in the field. Additionally, it provides a reference for the determination of the degree of filth in electric power patrols and laboratories.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.