Hydrophobicity classification of composite insulators based on convolutional neural networks

摘要

This paper discusses the classification of composite insulators in hydrophobicity classes, according to the spray method of IEC Standard 62073, using convolutional neural networks. By applying the spray method, about 4500 photos were collected and are available online, from all hydrophobicity classes using distilled water-ethyl alcohol as spraying solution. Convolutional neural networks were trained, validated and tested, in order to determine the hydrophobicity class of composite insulators and to eliminate the operator's subjectivity, which is the main problem in this measurement. Various configuration setups of convolutional neural networks are applied and compared for their appropriateness in accurately classifying the composite insulators. The proposed methodology is a useful tool for the classification of composite insulators in hydrophobicity classes restricting the subjectivity of human judgment. The experiments showed this method gives almost 98% accuracy in this classification task. Therefore, the proposed methodology is helpful in maintaining of composite insulators.