Improved Detection Of Fault Diagnosis In High Voltage Transmission Lines Using Thermal Imaging Based Convolutional Neural Network Module

Abstract

In recent decade, several efforts are made to increase the quality of transmitted power using faulty diagnostic techniques in high voltage equipment. Most of these techniques aims at discovering the symptoms of faults and in future it ensures that the fault does not occur in the same location. Such efforts are made in order to increase the efficiency of power transmission even if a faulty insulators exist. However, these techniques are affected by its own shortcoming that leads to high risk inspection and low reliability in monitoring, since the fault condition vary depending on the fixture and environment. Hence, to improve the reliability of monitoring and to prevent the accidents, the proposed system enables a Hardware and a Software Prediction module. The Hardware module consist of a mobile robot with a thermal imaging camera that moves along the transmission lines. The output from the Hardware module is sent to the software module, where the predictions are made prior the faults over transmission line based on image processing analysis. The software module uses Convolutional Neural Network (CNN), a deep learning module to predict the faults in advance based on the images collected by the thermal cameras. A threshold level based on outside conditions like temperature, humidity is set in CNN and it is allowed to train based on the existing fault imaging over the same transmission line. Further, the system is allowed to test in real-time using both hardware and software module. The predictions of the proposed system is compared with other existing methods in terms of detection accuracy and precision of faulty diagnosis. The result shows that the proposed method achieves improved rate of detection than the existing methods.