Abstract
Insulators are vital isolating tools in power transmission lines that usually function under harsh weather conditions. Undesirable operating conditions risk insulator surface contamination and consequential issues, such as leakage current and power outages. Therefore, the accurate condition assessment and maintenance of the high-voltage insulators remain important tasks. This paper attempts to use the frequency domain spectroscopy method for the first time in a high-voltage glass and porcelain insulator surface pollution severity analysis. To achieve this goal, 34 tests were conducted in two steps: pre-contamination and frequency domain spectroscopy measurements. First, a pollution solution was applied to the insulators, assigning each test to a specific pollution type. Next, the frequency domain spectra of these insulators were measured. The dielectric dissipation factor$(tan\delta)$I values for the different contamination levels were clearly different at very low frequencies ($10 \text{pHz}$to$100 \mu\text{Hz}$. The pollution level can be predicted with$0.004\pm 0.003$MSE for glass insulators and$0.002\pm 0.002$MSE for porcelain insulators using deep learning.