Classification of degraded polymer insulator using support vector machine

Abstract

The ability to monitor closely the surface degradation condition of polymer insulator will be really beneficial to the power utility company in order to ensure smooth and safe power transmitted to the consumer. If the level of degradation condition could be classified, then it could ease the maintenance team to take proper action as to avoid any undesirable event from happening. In this study, it has implemented the leakage current signal parameters data in the classification process of degraded field-aged insulator. These signal parameters are extracted from the Spectrogram. Prior to this analysis, the leakage current signal is captured during the testing method of inclined plane tracking. The physical evaluations such as arithmetical mean of surface roughness and static contact angle are also measured for the purpose of comparison of surface conditions. The Support Vector Machine is implemented in the machine learning test, in which the percentage of classification accuracy between degraded sample and the controlled sample is recorded. To validate the classification results obtained, the insulator sheds under test was going through the Spray Method to determine the criteria of hydrophobicity class in Table 1 of the IEC TS 62073:2016. By using the percentage of total harmonic distortion data, the consistency results of the classification accuracy percentage have been successfully determined the two significant classes and the transition class between them. However, there is an existence of insignificant classes if the root means squared leakage current data is implemented. Therefore, by implementing the appropriate leakage current signal parameter data, the degradation classification could be determined accurately.