Characterising electrical trees in nanocomposite epoxy-resin using partial discharge detection methods

Abstract

Modern polymer based solid electrical insulation technologies, whilst highly optimal for their function, have been shown to have a major aw: suscep- tibility to electrical treeing failure mode. This dissertation investigates the possibility of using Partial Discharge (PD) detection techniques to monitor and characterise electrical trees in nanocomposite epoxy insulation. The ex- perimentation is done by creating highly divergent electric stress in nanocom- posite samples and monitoring PD magnitudes over time until failure. Fur- thermore, a remote monitoring system was designed to allow for constant monitoring from anywhere with an internet connection. The experimen- tal results revealed several interesting ndings. Electrical trees in un lled insulation behave di erently from those in nanocomposite insulation. The electrical trees that form in un lled epoxy are of bush/branch type with the corresponding PDs comprising of both big (greater than 10 pC) and small (less than 10 pC) pulses throughout the lifetime of the tree. In nanodielectrics however, the PD magnitudes are generally smaller than in the un lled dielec- tric. Moreover the PD patterns evolve through distinct phases as the trees propagate from inception to complete failure. These distinct phases show new behaviour not seen in un lled insulation, with phases of complete si- lence and some displaying patterns similar to corona. The resulting patterns are analysed and auto-classi ed, using a program written for this purpose, before being used to develop a model to explain the observed behaviour.