Visual Recognition of Electrical Discharges by Video Processing

Electrical discharge has many fractal characteristics such as its sound and light besides from its electrical properties. In this study, partial discharge recognition is carried out by its visual light characteristics. Partial Discharge (PD) is an electrical discharge that is seen only at a part of the insulator between two conducting electrodes without completely bridging the gap. Partial discharge can occur in voids or liquid suspension in solid insulation, along the interfaces of multi-layer solid insulation systems, in gas bubbles in liquid insulation or around an electrode in a gas. Partial discharge causes deterioration and may lead to a complete breakdown of the system in time. Since partial discharge might be very destructive to power systems, it is very important to understand the mechanism of partial discharge. In this study, a technique for visual recognition of partial discharges by video processing is introduced. Required data for the study is obtained experimentally, because PDs are highly complex and it is difficult to study analytically. A measuring system; a regular camera to capture images along with the high voltage measurement apparatus, is installed for experimental study. High AC voltage is applied to two different electrode configurations. PD is intentionally generated by using rod-plane electrode system and recorded by the camera. Acquired data is then used for recognition of PD by using video processing techniques. The correlation between visual images of PD and applied voltage during a discharge is investigated. The effect of electrode separation on PD images is also examined. Different electrode systems are designed by using two different profiles for the rod electrode. These configurations are tested and, PD images are captured by regular camera and stored in order to carry out some analysis by video processing to get more information on PD phenomena. As a result of video processing, informative data about PDs such as propagation and light intensity of PD can be achieved. The possibility of detection, classification, diagnostics and monitoring by video processing for PD studies is investigated in this thesis. Therefore, by understanding the discharge mechanism, identification, quantification and risk assessment of insulation defects in power system will be feasible. Keywords: Partial discharge, corona, visual recognition, video processing.