Developing models to improve the online measurement of water in a vegetable oil filled power transformer

The existing online methods to monitor the water content of transformer insulation use water activity probes suspended in the oil. Mathematical models, based on cellulose adsorption isotherms, are required to calculate the water content of the solid insulation of a transformer. However, a disadvantage in using these models is the significant level of noise caused by changing temperature. Digital filtering techniques are being used to remove this noise, and thus improve the determination of water content. When designing a digital filter its delay and order must be taken into account. As the order of a filter increases, its transition band is reduced, however, its delay also increases. In this article a technique is proposed to improve the performance of a digital filter. A model to determine how the water content of pressboard varies due to a measured temperature change is presented, and this model is used with the digital filter to determine the change in water content. The filter was investigated using measurement data from an operating power transformer.