| Popis: |
Insulators play a vital role in electrical power transmission systems. Insulators are crucial components in serving to support the conductors and provide electrical insulation between the conductors and the supporting structures, such as transmission towers. Insulators are always exposed to vagaries of environmental conditions like rain, wind, snow, icing, and external over-voltages, causing flashovers and breakage, which leads to flaws/damage to the insulator. It is necessary to monitor the condition of the insulator strings of the power transmission line to detect the insulator's flaws early for safe operation. The proposed method uses a transfer learning-based EfficientNet model for insulator flaw detection. The inputs used are the images of the insulators collected from IEEE data port. The features are first extracted from the insulator images that are given to transfer the learning-based EfficientNet model to detect if the insulator is flawed or not. A comparative study of different EfficientNet models, such as EfficientNetB0, EfficientNetB1, EfficientNetB3, and EfficientNetB5, has been carried out. The accuracy of the method is found to be 99.9 % for both EfficientNetB3 and EfficientNetB5. EfficientNetB3 may be chosen as the optimal network owing to its high accuracy and less complex structure. The average precision of the method has been increased if compared with the traditional methods of insulator flaw detection. The main novelty of the proposed method is that it is the first time an EfficientNet method has been employed for insulator flaw detection and it provides high detection accuracy of 99.9 %. The proposed method can be used for accurately detecting the flawed insulator in power transmission lines. |
