Flashover risk prediction on polluted insulators strings of high voltage transmission lines

Partial discharges on high voltage insulator surfaces are directly related with the deposition of pollution over the insulators. A complete partial discharge sensor network was previously developed and has been in operation for approximately one year. A fuzzy inference system was applied on the results to extract a risk level. In this paper the risk level signal was applied as a neural network input to predict the flashover risk ten days ahead. Two sensors installed on transmission towers submitted to match environments provide the data sets used as training and test. The risk predicted follows the behavior of the observed risk and could be used to better plan the schedules for preventive maintenance.