The activity of partial discharge (PD) is one of the major causes of failure of high voltage transformers. Partial discharge detection is usually used as a preventive technique to avoid large-scale losses in the power system. Suitable and reliable methods to detect and localize PD in equipment winding insulation, such as in a transformer, are still widely researched. Among the methods used are based on the acoustic wave, on the electrical determination of the dominant capacitive region in frequency domain, and on the travelling wave method. After a brief review on current methods, a technique based on the frequency spectra of the measured signals showing the characteristic of the transfer functions is proposed. A Matlab simulation of the proposed technique is then presented. PD signals are injected at various points in a transformer winding to represent PD sources at different location within the winding. The main difficulties encountered in current methods, which is the noise and the limit in the detection frequency range, is overcome. In the frequency spectra of the proposed method, the crests and troughs could be used for locating the source of the discharge activity in a power transformer winding.
