An advanced model-based diagnosis system for online detection of the moisture content of power transformer insulations

Abstract

It is a well known fact that the moisture content of a power transformer insulation system is a key parameter for the estimation of its aging condition and operation reliability. Therefore detection of the moisture content is a very essential task within power transformer diagnostics. While conventional methods for estimation of the water content of the insulation system require a transformer shutdown for several hours, the proposed method calculates the profiles of water concentration in the cellulosic part of the insulation system, as well as the concentration of water in transformer oil from operational parameters. An advanced model for the moisture diffusion in paper estimates the temperature dependent moisture distributions in the paper for non-uniform temperature distributions along the transformer winding. The model covers temperature dependent diffusion coefficients and acts as a non-linear control loop for the water concentration in transformer oil. Moreover the water concentration at the oil-paper interface, which is a measure for the risk of surface discharges, is computed. The identification procedure for the models is based on nameplate data and dielectric response measurements. Sensors for temperatures and relative water concentration of oil provide the possibility to integrate the models into a framework of model-based diagnosis using structured residuals. The models have been verified with a laboratory setup consisting of a power transformer, equipped with sensors for temperature, load current and oil humidity.