Abstract
Previous models have examined the relationship between insulator pollution level and flashover voltage, as well as between pollution levels and image-based features. This study investigated the potential correlation between visible-light images and flashover voltage. Porcelain cap-and-pin insulators were contaminated using a modified fan-shaped pollution distribution technique with a low-volume medium-pressure spray gun. Four pollution levels were produced by varying the ash–salt concentration (0.02–0.12 g/mL) and the number of spray gun trigger actuation (2-12). After ultrasonic humidification at 90% relative humidity for 90 min, the AC flashover tests were repeated at each pollution level until ten flashover voltages were obtained. Insulator images were captured under controlled conditions, including illumination (6,000 lm, 6,500 K), exposure parameters (ISO 500, f/16, 1/10 s), imaging geometry (1.3 m distance, 11.5° downward tilt, maximum optical zoom), and in-camera sRGB color space. Image processing involved selecting the fan-shaped area, cropping, background removal, conversion to RGB and HSV stacks, and histogram computation. Pearson’s correlation analysis showed that the histogram peak displacement between the dry and wet conditions of the pollutant layer was the most representative image feature with respect to flashover voltage. The RGB histogram peak displacement exhibited a higher correlation level than HSV, with the red component showing the greatest correlation coefficient (0.82). Furthermore, multivariable linear regression based on RGB histogram peak displacement yielded a coefficient of determination of 0.915. These results indicate that visible-light images can serve as non-contact predictors of the insulator pollution level and flashover voltage.
Funding Information
  • Indonesia Endowment Fund for Education (LPDP) Doctoral Scholarship, Ministry of Finance, Republic of Indonesia
  • Universitas Gadjah Mada

This publication has 46 references indexed in Scilit: