Convolutional Neural Network to Detect Support and Suspend Insulators on Infrared Images
A. D. Zaripova · 2023
The thermographic evaluation method is widely uses in substations to detect electrical equipment and their faults. A lot of infrared images for each substation are analyzing by humans. To automate monitoring insulation, this paper proposes a convolutional neural network, which detects different supporting and suspending insulators from infrared images f. The technique may be used for online monitoring insulators in substations. The neural network model was tested on 2000 infrared images. Before testing the neural network, images were created from 900 infrared images of insulators belonging to 2 classes for training neural network using data augmentation methods. For evaluating the neural network model, were used statistical characteristics like metrics and loss functions to show minimizing and optimizing parameters of neural networks (weights) during the training. Results demonstrates that detection of support and suspended insulators reaches 98.81%. As a result, this proposed method with evaluated neural network can be used in monitoring systems to detect insulators before detecting their faults.