Enhanced partial discharge location determination for transformer insulating oils considering allocations and uncertainties of acoustic measurements

Abstract

The consequence of partial discharge (PD) activities inside transformer may result in a catastrophic failure. The exact PD location is based on precisely computation of the time difference of arrival (TDOA) between the signals at different acoustic sensors, therefore, the generalized cross-correlation approach is used to accurately determine TDOA between the sensor signals. Reducing the sensor location errors is necessary to identify the exact PD source location. 87 PD fault locations with 13 suggested sensor’s locations are presented to determine the best location of the sensors. The best sensor’s locations are determined based on the behavior of the maximum and minimum errors for each sensor’s locations. A proposed ANN model is constructed with different uncertainties of TDOA measurements and estimations. The ANN model is constructed based on 15,877 PD locations for each 0%, 5%, 10%, 15%, and 20% uncertainty noise of TDOA based on the optimal sensor’s location. Experimental works are carried out to verify the robustness of the ANN model. The maximum error of ANN model to determine the exact PD location is 2.74 cm with 20% uncertainty noise. On the other hand, the maximum error of other literature models is about 7 cm with the same uncertainty noises.

Keywords

• Power transformers
• Partial discharge
• Time difference of arrival
• Artificial neural network