Sensitivity of estimated parameters in transformer thermal modeling

In this paper a model for predicting hottest-spot temperature in the windings of a transformer is presented. Three forms of the model that can be fitted to measured field data are derived. The three forms of the model are then fitted to measured data and the model parameters are estimated. Of the parameters estimated, the time constant, we find is inconsistently estimated across the phases of the transformers while the ultimate hot-spot rise over top-oil temperature at rated load is estimated consistently. It is shown, through sensitivity analysis of the model parameters and graphical plots, that the inconsistency in the estimation of the time-constant parameter is caused by sensitivity of the time-constant parameter to noise in the input data.