Wideband parametric identification of a power transformer

This paper applies a recently developed technique for wideband frequency domain system identification to parametric modelling of a power transformer using frequency response data. It is well known that frequency response data from a typical power transformer contains many resonant modes and spans several decades in frequency. This in effect leads to a numerically ill-conditioned problem when attempting to fit a parametric model to the data. The method employed here to fit the parametric model utilises what has been termed frequency localising basis functions. These functions have been shown to improve the numerical conditioning of a least squares type estimator which correspondingly results in an increased accuracy of the estimated parameters. Two examples are used in this paper to highlight the veracity of the frequency localising basis functions technique. The first example uses simulation data, whilst the second utilises frequency response data from a large power transformer.