A new method for the estimation of time difference of arrival for localization of partial discharge sources using acoustic detection technique

In an acoustic partial discharge (PD) detection system, estimation of time difference of arrival (TDOA) between acoustic signals arriving at a sensor array is an important criterion for accurate localization of PD sources inside a transformer. The localization accuracy can be improved by improving the accuracy of estimation of TDOA between sensors. The estimation of TDOA is a challenging task because acoustic signals are corrupted by noise, reverberation, echo and reflection of acoustic signals inside the transformer tank. Keeping this in mind, this paper presents a technique for the accurate estimation of TDOA by extraction of an estimate of the PD pulse from the recorded acoustic signals. The TDOA between two sensors is then calculated by finding the cross-correlation function between the two sensors. The acoustic path through the transformer tank and oil constitutes the physical system, which when excited by the PD pulse, gives rise to the acoustic pressure waves. An estimate of the PD pulse, which generates the acoustic pressure waves, may therefore be obtained by separating the acoustic response of the tank-oil physical system from the acoustic signal. The extracted PD pulse information gives an estimate of the instant of appearance of the PD pulse at each sensor, which makes the accurate estimation of TDOA possible.