| To provide for the reliable operation of large power transformers, a dielectric coolant, transformer oil, is circulated through the winding and spacer structure to remove the heat of electrical losses. This gives rise to the sometimes catastrophic phenomenon of streaming electrification. Many of the failures resulting from streaming electrification occurred within a few hours of the initiation of flow, indicating that startup transients of streaming electrification may significantly affect the system. This hypothesis was investigated in an experimental flow system. The results, reported in this thesis, reveal the existence of two transients, a short term transient that is re-initiated each time the flow is initiated, and a long term transient that re-initiates only with the renewal of the interface. The short term transient is related to the re-distribution of charge in the double layer, while the long-term transient has to do with the charge adsorbed on the interface and the products of electrolysis due to the leakage current across the interface. The short term transient was further investigated using a finite difference analysis of the charge transport equations. The computational model provided information not only on the velocity dependence of steady state and transient characteristics, but also on geometry, conductivity, diffusivity, and electric field dependence. A number of mitigation efforts are recommended, including the use of a novel intelligent controller for the power transformer cooling system, which uses a combination of fuzzy logic and neural networks to operate the transformer in temperature and flow regimes that pose minimal risk of streaming electrification. |
