Detection of factor that can make failure in energy transmission lines with unmanned aerial vehicle and thermal camera by artificial neural networks method

In our developing world, the demand for energy is constantly increasing. In order to meet these demands, new connection points have been included in the electricity distribution network. The exploration of all equipment and facilities that make up the expanding electricity distribution network should be completed within two years. Types and planning of maintenance works vary in line with exploration activities. Exploration with traditional methods causes high costs, loss of workforce, lack of detection and risks to life. The factors causing the most failures compared to previous years were determined and images were actively collected from the field. The classes of these images; tree contact to the lines, bird's nest contact, presence of foreign objects, the concrete pole and sleeper to have completed their economic life and the absence of original fuses. Model design was carried out in convolutional neural networks according to the state of the data set. The integration of this model with the unmanned aerial vehicle with Python language is provided with the Pycharm interface. In the error matrix, it is seen that the original fuse was not used in the determination with the highest F1 score with a success rate of 97% in 5 different classes.