Danger assessment of the partial discharges temporal evolution on a polluted insulator using UHF measurement and deep learning

Pollution over insulators surfaces in outdoor environments is detrimental for long term operation of power systems. The temporal evolution of measurement signals as the contamination increases has not been given much attention. This work proposes presented the analysis of the time series of partial discharges measured with an antenna in an increasing pollution condition until flashover, using a deep learning algorithm in order to identify the early signs of an incoming flashover. Flashover was produced by gradually increasing pollution over a bushing insulator to carry out a binary classification of signals as low or high danger. Different time thresholds were tested and it was concluded that partial discharges measured with antennas can be used as early detection of flashover, and a time threshold at the 70% of total experiment time gave the best result, being noticeable transition from low to high danger signals before flashover.