Corrosive sulphur condition monitoring of oil-filled power transformers

The after-effects of corrosive sulphur-related faults on power transformers are disastrous. The existence of sulphur corrosion in transformer insulating oils leads to a formation and deposition of copper sulphide, creating a low resistance path across and through the paper insulation, resulting in internal discharges, flashovers, and ultimately catastrophic transformer failures. The current techniques applied to detect and monitor the progression of sulphur corrosion in transformer insulating oils (i.e. X-ray fluorescence and scanning electron microscopy–energy dispersive X-ray spectroscopy) require relatively expensive equipment and skilled personnel to run the tests, which increases operation costs. In addition, this approach must be carried out in laboratory, which requires a longer time to obtain the results. The existing standard tests, such as the BS EN 62535, ASTM D1275-B, and ASTM copper strip corrosion standard are imprecise because they are based on visual comparison of metal strips against a standard reference colour chart. As a solution, the work presented in this thesis is focused on developing a low-cost, portable, foolproof, and reliable method that provides instantaneous response in detecting the progression of corrosive sulphur compounds present in transformer insulating oils. In particular, the techniques investigated in this study were ultraviolet–visible spectroscopy, direct current conductivity measurements, and interdigitated capacitive sensors. The results reveal that these techniques were not feasible for the detection and condition monitoring of corrosive sulphur. However, obtained results and analysis led to the development of a novel technique based on thin film sacrificial copper strip sensors. It is possible to continuously quantify the elemental sulphur concentrations in transformer insulating oils instantly on-site and has potential to be implemented as an online condition monitoring tool. The design and development of the sensors were carried out based on the 22 factorial design. A regression model was developed, being able to predict the transformed resistance of the sensors as a function of their area and thickness. In use, the sensor change in resistance is a function of the level of corrosive sulphur compounds (i.e. in ppm) in the mineral oil