An Advanced Detection Method for Assessing Aging State of Composite Insulators Under Pollution Cover

Abstract

Ultrahigh voltage direct current (UHVDC) converter stations are crucial components of future power systems. However, composite post insulators, which are essential to the functionality of those stations, are susceptible to issues such as contamination deposition and aging due to exposure to ultraviolet radiation and atmospheric pollution. These problems can lead to flashover, reduced hydrophobicity, increased leakage current (LC), and pose operational risks. This article utilizes hyperspectral imaging (HSI) technology to detect insulator aging under polluted conditions. By analyzing the spectral similarities of insulators before and after cleaning, a spectral difference model is developed, extracting 113 characteristic bands indicative of aging. To reduce complexity, principal component analysis (PCA) is applied for dimensionality reduction, with PC1 (96.54% contribution) used as the dataset. The least squares support vector machine (LSSVM) algorithm, adapted for large-scale data processing, is employed as the classification model and optimized using improved Harris hawks optimization (IHHO), which identifies optimal hyperparameters ( ${C} =257.002$ , ${g} = 0.141$ ). The performance of the PCA-IHHO-LSSVM model is evaluated against five other classification algorithms, showing the best accuracy (98.07%) and computation time (1.935 s). The model is then used for visual assessment of insulator aging states, contributing to the identification of operational risks and ensuring the safe and stable transmission of energy in UHVDC converter stations.