Modeling leakage current in naturally contaminated insulators using linear regression for transmission line diagnostics

Natural contamination can significantly influence the surface conductivity of high-voltage ceramic insulators and increase the risk of flashover. This study proposes a modeling framework based on multivariate linear regression to characterize leakage current behavior induced by three contamination types: dust, moss, and flashover residues. Experimental data were obtained from controlled laboratory tests with three replications for each contamination type. Total Harmonic Distortion (THD), Equivalent Salt Deposit Density (ESDD), Non-Soluble Deposit Density (NSDD), temperature, and humidity were incorporated as input parameters. THD and NSDD were found to be the dominant predictors for dust-contaminated insulators, with a highly accurate model (prediction error < 1%). For moss contamination, the primary contributing factor was NSDD due to its hygroscopic nature, while THD served as a secondary indicator of electrical distortion. In the case of flashover residues, NSDD exhibited a strong positive contribution to leakage current, while ESDD showed a negative correlation, indicating the non-uniform and degraded nature of post-flashover surfaces. Model validation demonstrated strong agreement with experimental results. Prediction errors were generally below 1%, confirming the robustness of the proposed approach. Among all contamination types, NSDD consistently appeared as the most influential parameter, while THD effectively captured waveform distortion associated with surface degradation. The novel integration of FFT-based harmonic analysis with regression modeling proposed in this study enables a more comprehensive interpretation of electrical and physical interactions on contaminated insulator surfaces. The proposed models provide a practical and cost-effective diagnostic tool for monitoring transmission lines, supporting condition-based maintenance strategies, and enhancing system reliability in tropical regions prone to contamination.