Automated Identification for High Voltage Isolation Switch via Generative Adversarial Network

The detection and recognition for different states of high voltage isolation switch plays a key role in the quality management process of the electrical power systems. In the literature, most of the related techniques focus on the classification of deterministic states and ignore the identification of non-deterministic ones. Bearing this in mind, we propose a generative adversarial network (GAN) based approach to detect and recognize various cases of the state of high voltage isolation switch. To note that the presented GAN is supposed to increase the number of images since insufficient-sample is frequent in a majority of machine vision-oriented tasks. Meanwhile, we manually collected the data required for the experiment to build a data set, and compared the experiment with the most advanced technology methods. The final result shows that the superiority of our presented method in classification accuracy.