Classification of pollution severity on insulator model using Recurrence Quantification Analysis

In this work, a novel approach is established in order to investigate and monitor the performance of high voltage insulators. Since leakage current (LC) waveforms are intimately linked to pollution severity, it is primordial to study and investigate leakage current characteristics during the entire contamination process. In this paper, performance of a plane model insulator is studied through a number of laboratory tests under various levels of pollution contamination. LC waveforms are investigated through a nonlinear method called “Recurrence Quantification Analysis” (RQA). This method revealed successfully the non-linear characteristics of LC for identifying the dynamic behaviors on the insulator surface. Moreover, RQA indicators are found to be directly linked to the contamination severity. Thus, mean values of these indicators are computed and used as an input to three different classification algorithms (k-nearest neighbors, Naïve Bayes, Support Vector Machines) in order to classify contamination severity.